✨│ai-help

[timid valve]

why would you need to split them up?

[pastel oak]

delete the file "stored_settings.json" from the folder, and try again

[viral crater]

in powers of ten to be exact

[viral crater]

trying rn

[viral crater]

thx

[odd shale]

Because RVC has already an built-in audio splitter, but it's kinda dumb and not reliable.

So instead i suggest you to split your .wav files on clips of 10 or 5 seconds

[timid valve]

does that have an effect on quality?

[odd shale]

Kinda, but it's better to avoid them. But if you get mode collapses anyway it will depend on your dataset's quality.

It helps to avoid mode collapses.

[timid valve]

oh, okay... what are those?

[flint geyser]

basically when the Adversarial network freaks out and the whole training process breaks

[odd shale]

Yep, that's true.

[flint geyser]

the adversarial network is a part of the training

it checks how close the current model is to the source material

[odd shale]

Thanks for explaining it buddy.

[flint geyser]

no prob bob :D

I'm sorry i was inactive for a long time, i was gone litteraly only for school and vacation

[timid valve]

all good man, glad to have you back

[viral crater]

@pastel oak its very laggy idk how to fix it

[pastel oak]

What does it say on "GPU : " ?

[flint geyser]

Are you using the realtime voice changer? (w-okada)?

[viral crater]

cpu

[flint geyser]

AI is very laggy on cpu only

[pastel oak]

Is your gpu showing up?

[viral crater]

no

[flint geyser]

try using cloud-based options

[viral crater]

does that mean i dont have a gpu?

[pastel oak]

Dang i really dont know what you did for your gpu not to show up. do you really not have a gpu? lol

[flint geyser]

are you on a laptop?

@viral crater

[viral crater]

ig not

[rare goblet]

Ayo? @viral crater level 3 !!!

[viral crater]

no im on desktop

[flint geyser]

hmm

are you sure you have a GPU installed on your computer?

[viral crater]

no

i dont think i have one

[flint geyser]

Try cloud based options for RVC then.

[grave mortar]

[viral crater]

cloud based options?

[flint geyser]

yep

[grave mortar]

-cloud

[flint geyser]

like https://docs.aihub.wtf/rvc/cloud/applio-colab/

you can find other options in the guide

[flint geyser]

What seems to be your problem?

[viral crater]

google colab?

[flint geyser]

yip

[viral crater]

i tried that millions of times

[flint geyser]

these are up-to-date

[viral crater]

never worked once

[pastel oak]

l3af hold on a sec

[flint geyser]

alr

[pastel oak]

hes talking about realtime voice changer not this haha

[flint geyser]

oh lol

nvm

[pastel oak]

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

The guide on setting it up:
https://docs.google.com/document/d/e/2PACX-1vTIceEcBfS6Zqolv_QEysrFfI_EJikPxozWptP_EjkpLVl-l-gdo-ijBonQMTviAHEYm5emmd9k9TdC/pub

Already a heads up, you can skip the "[Optional] Upload a voice model (run this before the voice changer)" since it doesnt work, you can upload it later when the GUI opens

You can try running this

BUT

[flint geyser]

you should maybe think of investing a gpu for reliable realtime voice changing.

[pastel oak]

this relies on your internet, and since you struggled to download 5gb earlier i really dont know if you will have great results with this

[flint geyser]

@viral crater you should maybe think of investing in aa gpu for reliable realtime voice changing.

[viral crater]

so i basically cannot have good ai voice changing without gpu

[flint geyser]

yup

which country do you live in?

[viral crater]

canada

[flint geyser]

oh damn

you can find really cheap GPU's there

like spend $50 maybe

and thats it

MAKE SURE THE GPU IS COMPATIBLE WITH YOUR COMPUTER

and follow proper instructions or get a professional to do so

[viral crater]

wait my dads computers got a gpu 🤔

[flint geyser]

yeah

[viral crater]

ive got an idea!

takes gpu

[flint geyser]

you could ask him

[flint geyser]

lel

[brittle wing]

is there a live conversion thingy??

[flint geyser]

yup!

[proven hill]

#🔍│help-w-okada

[low shard]

realtime voice changer for calls?

[flint geyser]

you'll need a GPU for it

[flint geyser]

#🔍│help-w-okada for more info

[brittle wing]

odaka dont work for me 😭

[flint geyser]

do you have a GPU

[proven hill]

prob bad settings

[flint geyser]

and is it Nvidia

[low shard]

that depends on your GPU, do ctrl+shift+esc , performance tab and tell whats your GPU name

[brittle wing]

Intel Iris Plus Graphics

[flint geyser]

yup

you won't be able to run voice changer smoothly then

[brittle wing]

can i run it at all

[flint geyser]

yeah

but VERY laggy

[brittle wing]

thats fine

[flint geyser]

alright

[viral crater]

i will come back to this program when i have a gpu

[low shard]

yea that cant work

you have to do it on cloud

[flint geyser]

good idea

[low shard]

one sec lemme send

-rt

[azure marsh]

This interaction has expired, use the command /guides realtime if you wish to see it again.

[flint geyser]

ay nick can you send me a cheatsheet of automase commands

[proven hill]

i dont think you understand how laggy tho

[low shard]

Also
Kaggle (a bit harder to use and needs phone number but gives more hours than google colab)

Original Wokada version
Modified Wokada version

[low shard]

cheatsheet?

if you wanna know all commands of automaze

do

-help

[azure marsh]

Command has expired

[low shard]

and u can see all the commands of each category u select

@quasi dagger for the online realtime automaze options, could u:

• Modify: 'How to use W-Okada's Voice Changer' -> 'How to use Original W-Okada's Voice Changer Google Colab'
• Add: Modified W-Okada's Voice Changer Google Colab
• Add: Original W-Okada's Voice ChangerKaggle
• Add: Modified W-Okada's Voice Changer Kaggle

maybe also add that google colab gives 4 hours daily and is easy to use while kaggle gives 30 hours weekly of better gpus but its harder and needs phone number
Idk where you could write that tho as it should also be written on -colab and -kaggle too then 😭

[clever crater]

anybody got any tips on making your voice much more responsive in terms of seconds and less laggy? i cant tell if its my pc or it being laggy

[violet heron]

Can you send a screenshot your setttings?

[molten fog]

anyone have a good workflow for seperating singing voices without rx11?

currently just using bs roformer and then normalizing resampling and shit afterwards buti feel like that isnt good enough.

[thin stump]

BS Roformer > UVR Denoise > normalize > auburn noise gate> audio labeling. No guarantee its good because you can't inspect your audio closely in audacity

[molten fog]

should be decent at least though?
i was considering doing Mdx23c > Bs Roformer v2 > Foxjoy’s dereverb hq > UVR de echo > Uvr de noise

[rare goblet]

Ayo? @molten fog level 3 !!!

[thin stump]

You don't need MDX23C since BS Roformer is the best one. I also dm'd you

its a windows only app

just in case you're on mac

[lost sail]

-collab

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[molten fog]

also if i have a 3 minute file that is just seperated audio from bs roformer combined together then can i uvr denoise that whole file since instead of individually denoising each sample and combining them again

[thin stump]

yeah just combine files. I split my files into 3 minute sections if its more than 10 minutes long . I assume you're using mvsep?

was going to say that Finetuned BS Roformer = expensive rented GPUs to create btw

[steep mantle]

how do you make pht files?

[thin stump]

https://docs.aihub.wtf/essentials/how-to-make-voice-models/

[molten fog]

yea mvsep its a 3 minute file that is part of my dataset (im gonna combine it with a 10 minute one later) but i havent denoised anything inside that 3 minute combined file so i just need to denoise the whole thing at once before merging audio to make a complete dataset

[opaque spoke]

how do you stop or pause training your model?

[thin stump]

close out rvc, your preprocessing and feature extraction is already saved in the logs folder

[opaque spoke]

then how would you continue after?

[thin stump]

You can do it all at once if its below 100mb. You just have to check the file size before you split it up

[molten fog]

yea its only 30 mb flac so im good

[thin stump]

Use the same model name, target sample rate, batch size, pretrain, and save frequency to train the model again in the RVC GUI

do not rerun preprocessing or feature extraction

[opaque spoke]

oh so you just press train?

[thin stump]

yeah with those same settings

[opaque spoke]

ok

also am i using the right gpu?

[molten fog]

also do i need to do audacity remove dc offset on every single sample song before seperating the audio with bs roformer?

or am i supposed to od that after seperating all songs and combinging them in audacity

[thin stump]

im not sure with GPU's because i use kaggle mainline

[opaque spoke]

cause like i dont know if im using my integrated or dedicated

i want to use my rtx 4060

[thin stump]

dc offset should be during your normalization section. But I already gave you rx11 so you just have to play around with that if you want to get to the next level

[opaque spoke]

yea i think im using my integrated card, how can i restart all over?

do i just delete the log file

[molten fog]

if it bad if i removed dc offset on each sample song individually in audacity before putting it in mvsep and then combined all those in audacity and removed dc offset again

[thin stump]

Might produce little noise but im suuure its fine because its not an eq

its just to remove bottom line noise

[molten fog]

alright

[opaque spoke]

do i delete the log file to restart my training?

[thin stump]

https://tenor.com/view/billy-butcher-scorched-earth-scorched-earf-william-butcher-the-boys-gif-9293382775817207493

[molten fog]

is uvr denoise on mvsep or should i just use ultimate vocal remover for that and if so what aggression

[opaque spoke]

?

[thin stump]

its the same devs

[thin stump]

delete these files including your extract and preprocessing files

oh wait

idk

you just have to change the gpu

[opaque spoke]

but like idk if im supposed to put a 0 or 1 here

also was i supposed to say yes to the far right question

[molten fog]

this?

[thin stump]

thats new and isnt good

its UVR denoise

[opaque spoke]

does anyone know?

[molten fog]

am i blind or is there no uvr denoise on mvsep

[thin stump]

Unfortunately blind. It's called ultimate vocal remover

[molten fog]

should i keep default aggression

[thin stump]

3-5 https://rentry.co/RVC-dataset-RX11#best-model-for-dereverb-and-de-echo

[wanton bane]

how do i remove echo from my vocal ?

[opaque spoke]

am i supposed to use these pth files to continue training?

[tame mica]

yeh

[opaque spoke]

do i put the full path?

or start at \logs

[knotty moth]

I'd suggest to enable "save only latest ckpt" so that it would be only D/G_2333333.pth

[opaque spoke]

what oes that do?

can i change that setting in the middle of it

[knotty moth]

before starting training

[opaque spoke]

what do i put?

full path or \logs?

[rare goblet]

Ayo? @opaque spoke level 5 !!!

[wanton bane]

how do i remove echo from things like this sample ?

[knotty moth]

leave it as is unless you want to use a pretrain from #1235952130855010365

[opaque spoke]

oh ok so it will automatically start from where it left off?

[knotty moth]

if there are D and G files in the model's logs folder, it will resume from it

[opaque spoke]

will it be ok to change settings before continuing to run it again?

before i had all of them on no

[final sentinel]

I just tried uvr, and I was able to do bsrofomer, but I couldn't do all the vrartch models, so what should I do in this case?

[knotty moth]

only the "save only latest" option and more number of epochs if u wish

[opaque spoke]

so i leave the far right one as no?

[knotty moth]

you should use the same batch size and other settings as before, otherwise it would be messed up

[opaque spoke]

so i can only change this one?

[knotty moth]

better always enable it for most cases

[opaque spoke]

but like can i change it?

at the start i had it on no

[knotty moth]

I see no point of rolling back to a previously saved progress (like resuming from epoch 50 when currently on epoch 100)

[opaque spoke]

?

so do I keep all the settings the same???

as when I started

[signal willow]

when training a model, should i have a bunch of individual audio clips or one large audio clip, combining all the smaller ones?

[knotty moth]

Im sure the guide says so https://docs.aihub.wtf/

[opaque spoke]

I was just kinda confused did you mean this as like the next time I train a model you suggest me to do this?

oh yea probably cause you didn't know I was resuming training right?

[knotty moth]

what

[opaque spoke]

im resuming the training after I crashed

[signal willow]

???

[knotty moth]

that seems obvious from your first screenshot

[molten marlin]

how to reduce some noise on wasapi?

[opaque spoke]

oh wait yeah

uhh

[opaque spoke]

so will it work if I change that setting to yes? becausue i had it on no when i started

[knotty moth]

go do it instead of asking more nonsense

[opaque spoke]

yea sorry i was just kinda confused

also im not sure if im supposed to put a 0 or a 1 for my gpu

[knotty moth]

I'd hear more of your questions if you encounter some errors and issues, not when you havent started it at all

[opaque spoke]

bruh my mouse froze ;-;

ok

ill ask if i something goes wrong

[opaque spoke]

Is it normal. I’ve crashed three times when training

?

I just crashed again

[knotty moth]

any error message or power outage, or kind of abrupt crash? (check some errors in windows event log)

[opaque spoke]

No

My mouse freezes like a few minutes before it happens though

And then there’s a blue screen with a 😦

[knotty moth]

check in task manager if there are some bloats or background processes eating some cpu usage

or dont tell me if you are using an intel 13/14th gen cpu that is reportedly failing

[opaque spoke]

I’m using an intel 13th gen

What’s wrong?

[knotty moth]

toms hardware, techpowerup, and many other tech articles have been talking about that intel raptor lake failure issue. Intel themselves have offered extended warranty and might soon face lawsuit regarding it.

[opaque spoke]

so what do i do?

but how come my cpu is making it crash? I thought you used the gpu for training

[knotty moth]

yeah thats a serious issue, many people have prob replaced it with a working amd cpu

[opaque spoke]

i have a laptop

[knotty moth]

I heard that issue also applies on intel 13th i7 & i9 laptops

[opaque spoke]

so do i just give up??

[limber onyx]

How to make my own models

[knotty moth]

try ask your vendor and they might think so

[opaque spoke]

can i just use the model that i trained so far

its like at 100 epochs

[limber onyx]

how to train my own model chat

[opaque spoke]

docs.aihub.wtf

[limber onyx]

thank yousir

[opaque spoke]

how do i find the pth file?

[knotty moth]

assets\weights folder

[opaque spoke]

ok thanks

[knotty moth]

it is not optimal due to being undertrained, but u can try resuming it in a kaggle notebook https://rentry.co/RVC-Mainline-Kaggle

[opaque spoke]

the thing im confused about is why its crashing cause the cpu was only at like 40%

[opaque spoke]

ok ill look at that

[knotty moth]

the issue is caused by permanent degradation by sustained cpu overvolting (not the user fault but kind of design flaw that applies to all the same product)

[opaque spoke]

yeah the model sounds pretty bad

😦

[limber onyx]

guys can i uhh use an entire audio

or it has to be 10 seconds each like old rvc and stuff

i havent used rvc in long

[thin stump]

A combined audio file to clean it up then https://rentry.co/RVC-dataset-RX11#noise-gating-and-audio-labeling

[signal willow]

i ended up doing multiple audios, will it be fine?

[thin stump]

Yup, ideally just let audacity do the splitting itself with audio labeling next time

it'll remove most of the silences

in your dataset because there can be noise below -96db etc

[signal willow]

well, i am doing a character from a tv show, and i managed to get a full episode with surround sound downloaded, so i got the talking full quality, free of any noise

will that be good?

[scenic stone]

hi i have a questiom. If i have find an ai model with all members of onr group together and i vant find each member model alone how i can use it ????

[rare goblet]

Ayo? @scenic stone level 1 !!!

[knotty moth]

take the center channel that mostly contains the voicelines, or if necessary separate it from some other sounds using bandit plus in mvsep

[limber onyx]

can someoneh elp me

[rare goblet]

Ayo? @limber onyx level 1 !!!

[limber onyx]

/content/Mangio-RVC-Fork/logs/josesitopenesito/3_feature768

Exception Traceback (most recent call last)
<ipython-input-12-86df4775937e> in <cell line: 30>()
29 listdir_res = list(os.listdir(feature_dir))
30 if len(listdir_res) == 0:
---> 31 raise Exception("No features exist for this model yet. Did you run Feature Extraction?")
32
33 try:

Exception: No features exist for this model yet. Did you run Feature Extraction?

i already ran feature extraction

triedw running again

but it wont work still

im using rvc v2

cloud

[knotty moth]

if that step didnt show anything preprocessed and said "no-feature-todo", you have either submitted an empty dataset or zipped the dataset with wrong structure

if that issue also persists, re-export the wav files using audacity

[limber onyx]

yeah it said no feature tod o

i did what the site told me

make a flac file

put into zip

put it into rcv disconnect

wait its supposed to be wav?

[knotty moth]

try wav format

[grim fulcrum]

Is rvc v2 better than w-okada?

[limber onyx]

hey i put the wav and put a new zip

but it does the old zip

ok i tried everything

i refreshed apge

did everything again

now put a wav

[thin stump]

you should just use kaggle instead. It's much easier if you get right into it https://rentry.co/RVC-Mainline-Kaggle

I don't really like the bugs that rvc disconnected has because it's either google colabs fault or its an error with the session

[knotty moth]

I have always been doing audio labeling and exporting as multiple wav files in audacity, and then put them in a folder with model name. no issues encountered in rvc disconnected and anything else. so not really a bug.

[thin stump]

That's how I always did it too. I had to switch accounts once in a while because preprocess worked for the alt account but not for the main one

oh wait. He didn't include .zip lol probably what happened to me so nvm

[tough fiber]

@wispy lodge

bro i trained 2 model with applio on my local

its just only saving last epoch

it didnt save other epochs

i set this every 5 epoch but

its just only saving last epoch idk why

[knotty moth]

you mean save only latest D and G files?

[tough fiber]

no i mean model saves

when i use paperspace few months ago it will save model pth files every 5 epoch

[flint geyser]

ty after my hiatus kinda forgor everything

-help

[azure marsh]

Command has expired

[low shard]

Yw

[knotty moth]

mainline RVC should have saved weight files every n epoch set, as well as the hina's kaggle notebook

[tough fiber]

im using applio

i dont wanna use kaggle or something cuz i dont wanna deal with ngrok thing

paperspace was good but its gone

[knotty moth]

Applio 3.0.9 (local) should also have done so

[tough fiber]

as you see

i set it every 5 epoch but its only there 11 epoch rn

[knotty moth]

show me the console output when saving ckpts

[tough fiber]

yea it saving g and d thing

but they arent model pth file right

its config

when i was use paperspace it was saving model pth files every 5 epoch

not d and g

[flint geyser]

Hey there, does anyone need help?

[knotty moth]

ah that overtraining detector, haven't used that shit tho

[tough fiber]

yea i need we trying to find out

[knotty moth]

it seems you have only enabled the overtraining detector, so it only saves that "optimal" epoch

[tough fiber]

so if i turn off that option it will fix?

[flint geyser]

in response to atri's advice, if you want every 5 epochs to be saved no matter what, disable the overtraining detector

[tough fiber]

okey ill try

[flint geyser]

was i helpful?

[wanton bane]

does anyone know how do i remove echo from this sample ?

[wanton bane]

can you help me ?

[flint geyser]

Alright, give me a moment.

-audio @wanton bane

[azure marsh]

Suggestions for @wanton bane

📚 Audio Guides

• Creating Datasets for RVC using iZotope RX11, by Cauthess
• Perfecting Audio Isolation on Low-End Rigs, by Litsa The Dancer and Faze Masta
• Gathering and Isolating Audio, by SCRFilms :snowflake:
• Vocal Mixing Tutorial, by Roomie

🛠️ Tools

Audio Separation/Isolation

• https://mvsep.com/
• https://multimedia.easeus.com/vocal-remover/
• MDX-NET Colab

More commands: -colab, -uvr, -karafan, -overtrain, -help

[flint geyser]

check these tools

Try reverb removal

[wanton bane]

tnx alot

[flint geyser]

no problem

[proud anvil]

Please throw off the neural network that works

[pastel oak]

is this not wrong folder?

go to weights folder

you are in logs folder, it doesnt have the models saved in there

[stone lintel]

hi im getting this error when trying to download tensorboard can anyone help?

File "C:\Users\Tense\AppData\Local\Programs\Python\Python311\Lib\site-packages\tensorboard_vendor\html5lib_trie_base.py", line 3, in <module>
from collections import Mapping
ImportError: cannot import name 'Mapping' from 'collections' (C:\Users\Tense\AppData\Local\Programs\Python\Python311\Lib\collections_init_.py)
Keeping the command prompt open...
Press any key to continue . . .

the file in that path exists so idk why this eror happening

[tough fiber]

weights folder doesnt exist

i was tried to search that folder but theres no weights

only log folder here

[tough fiber]

didnt work bro

[rare goblet]

Ayo? @tough fiber level 13 !!!

[tough fiber]

still same shit

[queen kraken]

under assets

[tough fiber]

i cant see any weights folder bro

[queen kraken]

oh I use mainline not applio 😦

[knotty moth]

UVR dereverb or BS roformer dereverb

[knotty moth]

the weight files go into the model's logs folder in Applio

[pastel oak]

AH dang

[tough fiber]

still theres no weights

[knotty moth]

also I'd suggest enabling "save only latest ckpt" so that it will only save as D/G_233333.pth (not related to where it still saves weight files every interval set)

[tough fiber]

yea i turned off maybe thats why i cant see but its different thing

ill turn on

everyone was using applio when i use mangio. now im using applio and everyone using mainline rvc wtf bro

[rare goblet]

Ayo? @stone lintel level 1 !!!

[wanton bane]

i found a way with uvr5 but tnx

[knotty moth]

you can also download BS roformer dereverb for UVR https://github.com/ZFTurbo/Music-Source-Separation-Training/issues/1#issuecomment-2229279531

[knotty moth]

and then use the config file from mine

[wanton bane]

yours can do better than this ?

[rare goblet]

Ayo? @wanton bane level 2 !!!

[knotty moth]

quality-wise, UVR dereverb is still better than that, but the dereverb model I showed to you should be better than the former two

[wanton bane]

i will try it
tnx

[stuck slate]

when i click start audio conversion in rvc gui it freezes

[pastel oak]

Make sure your input and output both have the same ending with (MME)

[stuck slate]

omg

thank you

[pastel oak]

I havent used rvc realtime in a while but any other questions you can check out this guide
https://rentry.co/RVCRealtimeGuide

[flint geyser]

hey there, im back

who needs help

[brittle wing]

how do i make the vc run on a gpu?? yesterday i could run it on tesla t4 but now it js says cpu and it doesnt work at all now

[rare goblet]

Ayo? @brittle wing level 1 !!!

[brittle wing]

me

[flint geyser]

you have a gpu now??

oh

is it on cloud?

colab?

[brittle wing]

colab

[brittle wing]

idk much about all of that stuff but yestaday it said it ran on tesla t4

[flint geyser]

uh

try running it on another account or wait 24 hours

its how colab keeps it free

[brittle wing]

like anotehr google account?

[flint geyser]

yup

[brittle wing]

ok ill try

[flint geyser]

alright

[brittle wing]

thanks a lot

[flint geyser]

no problem

[half brook]

The only thing is I am on mobile rn lol @flint geyser

I just open my computer

[rare goblet]

Ayo? @half brook level 1 !!!

[flint geyser]

you should try to get access to a computer

[half brook]

Yep I will install apple

[flint geyser]

appolio you mean?

make sure you have an NVIDIA gpu for a better experience, or use cloud options

[brittle wing]

the file uploader doesnt work for me

how do i fix it

[worn river]

How do I train with Google Drive dataset?

[low shard]

#📰│dev-updates message

[worn river]

thanks, even I can hope a issue fix, I need help about drive datasets

[plush compass]

When you create drum AI, do you extract drums from only one song? Is it okay to add multiple songs? Also, do I need to remove the reverb from the drums?

[odd shale]

If you would like to create a Drums AI model of a song, you must use the drum stem of that sole song.

Not other songs.

And yeah, you need to remove reverb from the drums i think

[plush compass]

Thx

[turbid pagoda]

Can anyone tell me where to find the index file after I trained a model on applio o can find it and it wont come up?

[odd shale]

You should have trained the index before starting training the model.

There's a button called "train index" you must click before clicking on "train model"

[turbid pagoda]

ok so i kind of did it backworks got it.

[misty elk]

-rvc

[azure marsh]

Documentation

AI HUB Docs

https://docs.aihub.wtf/

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• Full AI Voice Model Training Guide (local) by Christopher Villanueva
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[misty elk]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[ashen vale]

is rvc still the best voice model to use currently or did it get outclassed like sovits did? I havent used anything ai in like a year

[misty elk]

-hf

[azure marsh]

<:huggingface:1179800228946268270> Hugginface Spaces

• UVR5 UI, by Eddy and Ilaria Huggingface Spaces
• Ilaria RVC Zero, by Ilaria Huggingface Spaces
• RVC⚡ZERO, by r3gm Huggingface Spaces
• Applio, by IA Hispano 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, -colabs, -kaggle, -local, -overtraining, -realtime, -rvc, -help

[analog obsidian]

Rvc still the best that we can run in consumer pc, microsoft has valle-2 which is super realistic but they didn’t release to the public due to security reasons

I think elevenlabs also has their voice cloning ai but ive heard is not that great (paid ofc and it doesn’t run locally)

Tldr; rvc is the best voice cloning ai we can use

[odd shale]

I hope it's matter of time till a public version of Valle-2 appears XD

[supple inlet]

I have a question about RX11. I have RX 11 Elements and use it as a plugin in Audacity.
So, I'm wondering if Repair Assistant can't be used in Audacity?
No matter what I try, I can't get the audio to play when the Repair Assistant tab is open...

[analog obsidian]

They can add like an inaudible watermark like synth v dev did with their voice creation ai

[analog obsidian]

Yeah audacity does not behave well while using rx plugins, they never worked well for me inside audacity

Only audacity devs can fix that

[supple inlet]

Is upgrading to RX11 Standard or using a paid DAW (like Adobe Premiere Pro) the only option?

[analog obsidian]

rx11 is recommended because is the most easy to use, i have no idea how well the rest of daws are for rvc, im not that experienced in daw usage, sorry

[odd shale]

I use both RX10 and RX11 in extreme situations.

[charred briar]

-realtime

[azure marsh]

This interaction has expired, use the command /guides realtime if you wish to see it again.

[lethal orbit]

Hello my friend has a problem

Whenever he speaks he has a echo and its sounds alot cutty

The settings r this

[pastel oak]

Disable sup1, disable echo for now

Use extra 8192

The echo is caused because your microphone is picking up your headphones volume.

Possible fixes:
keep sup 2 activated
Move S.Threshold all the way to the right
Lower windows audio volume so headphones are not as loud and dont get picked up my mic

If it still echoes, enable echo and see if it helps

[lethal orbit]

oh alr

btw what does the threshold do?

[pastel oak]

It is a noise gate, so it only starts picking up audio when it reaches a certain volume

so on the far right, it wont pick up quiet audio (which could help when volume is also quiet)

[lethal orbit]

oh he put it all the way to the right and it still doesnt work

its alr ill msg tmrw and see if it works

[empty fjord]

I am using AI Voice Changer, but I cannot successfully convert the voice. Normally, "rvc" should appear in the upper left corner of the anime character image, but it does not appear on my side. Does anyone know why?

[rare goblet]

Ayo? @empty fjord level 1 !!!

[weary oasis]

it keeps giving me error none

[final sentinel]

I can't do this, what should I do?

[tidal trail]

output folder = /content/drive/MyDrive/yourmusicfolder

input folder = /content/drive/MyDrive/youraudiofolder

[final sentinel]

I've done that and both completed, but the important thing is that the file hasn't been added to google drive.

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

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

Advanced Settings

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

⠀

Made with love by Ilaria 💖

[knotty moth]

このサイトを利用することをお勧めします https://mvsep.com

[knotty moth]

nah, just name those folders "output" and "input" respectively

[tidal trail]

yep

[final sentinel]

いや自分はこのcolabが使いたいんですなんでできないか教えて欲しいです

[analog obsidian]

mvsepの方がいいよ。サイトにあるモデルの方がcolabのよりいいんだ。

[final sentinel]

ごめんなさいーこれやったけどあんま精度良くなくてこっちのcolabの方にしたいんです。

[odd shale]

Taki taki rumba.

[ocean lark]

Every time i tab out of the voice changer client, the res shoots up infinitely. any way to keep this from happening?

[knotty moth]

mvsep のこのモデルの方が優れていると思います。colab のモデルよりも SDR が高くなっています。

[knotty moth]

このSDRリーダーボードを見てください

場合によっては、SDR が高くても必ずしも良い結果が得られるとは限りません。

[final sentinel]

それでもいいのでこのcolab教えてくれませんか？自分ずっとこれのuvr使ってきたんです

[knotty moth]

これはローカル UVR に似ていますが、colab ポートにあります。
たとえば、私が PC でゲームをプレイしているのに、あなたが Switch でプレイしていて、何か問題が発生した場合、私には助けることができません。

[final sentinel]

それでもいいです

[rare goblet]

Ayo? @final sentinel level 6 !!!

[azure marsh]

Command has expired

[loud sparrow]

#1175430844685484042 Assala Nasri

[rare goblet]

Ayo? @loud sparrow level 1 !!!

[violet heron]

You go into that channel and search it

[azure nimbus]

Realtime Voice Changer AI

any one have the 1.5.3.8a

version

[grizzled heart]

what is the app for making ai sounds?

[pastel oak]

8a is old, do you mean 18a?

https://rentry.co/VoiceChangerGuide

[pastel oak]

Do you mean AI Covers?

[grizzled heart]

yeah

[azure nimbus]

a8

[rare goblet]

Ayo? @azure nimbus level 1 !!!

[scenic arch]

stable audio unless you mean something else

[pastel oak]

https://docs.aihub.wtf/essentials/how-to-make-ai-cover/

If you have a good gpu, i recommend Local > Mainline
If you want to do it online fast, use this: https://huggingface.co/spaces/TheStinger/Ilaria_RVC

[pastel oak]

dont use 8a. Why are you looking for an old version?

[azure nimbus]

or 10a

[grizzled heart]

aight thanks

[rare goblet]

Ayo? @grizzled heart level 1 !!!

[pastel oak]

Why old version

[azure nimbus]

dm me

[pastel oak]

No

[azure nimbus]

i want to show you i cant send

[pastel oak]

Oke

[primal fulcrum]

Does anyone know how to make your own moduls?Cuz I tried this tutorial https://youtu.be/Evs8Or2Qy9A?si=vYUQb_Z2T7ClrZ1S but I didn'T have a rvc interface

[low shard]

That's a Google Colab YouTube tutorial, most YouTube tuts are outdated so it's better you don't follow that

What's your PC GPU?

[agile gale]

github download is slow any altnernatives?

[primal fulcrum]

NVIDIA GeForce GTX 1070

[primal fulcrum]

which methode works ? @low shard

[rare goblet]

Ayo? @primal fulcrum level 1 !!!

[knotty moth]

if you want to train locally, read the guide https://docs.aihub.wtf/
1070 is still capable of that despite being unoptimized

[rotund prawn]

Hi

[stone lintel]

hi quick question is uvr used for seperating vocals from music only or can it remove other background noises

[pastel oak]

Depends on what kind of background noises you mean, but yes it can

[stone lintel]

i guess it would be noises from the surrounding environment idk how to explain rlly

[pastel oak]

I havent opened UVR in a while but there are models called DeNoise. Someone else can probably give you better suggestions

[long matrix]

I can't run pip commands despite having the newest version of python installed? any ideas?

[primal fulcrum]

but how a get the voice from a audio to the pth data?

[pastel oak]

Please delete this if it contains copyrighted instrumental

[shrewd bolt]

ig they wont dmca u but okay

[pastel oak]

Only post the acapella of the model which you seem to be asking about

[shrewd bolt]

can u help me

[pastel oak]

Server got taken down before for this reason

[shrewd bolt]

they are kpop idols they wont dmca it

why my model sound too bad

im use pre train

[pastel oak]

Can be anything:

• your voice models dataset contained too much noise
• the audio file you inferenced (ai covered) with was not clean enough

[shrewd bolt]

let me send dataset

[pastel oak]

Did you send a small part to show me or is this the full dataset the 25 seconds

[shrewd bolt]

full dataset im need her voice only from that song

[pastel oak]

25 seconds dataset is tough i dont think you can get a good voice out of that

[shrewd bolt]

im used dolby atmos ver

[pastel oak]

10-15 minutes are recommended

[shrewd bolt]

okay

[pastel oak]

But training process is wild even 1 minute can work, maybe it didnt work for you in this case unfortunately

Post the thing you sent except without the instrumental if its really because of that, someone else will be able to identify this better than I can 👍

[primal fulcrum]

but how I get a voice to a pth data

[long matrix]

I can't find any RVC clients that actually let me download for realtime mic output..

[pastel oak]

You train it, read the how to make voice models part

Download the program under RVC > Local > Mainline

[drowsy scaffold]

hey guys um the software isnt installing how can i

[pastel oak]

Realtime voice changer
https://rentry.co/ForkVoiceChangerGuide

[drowsy scaffold]

its says too many people have installed it

and i must wait 24hours but

im not tryna wait that

[pastel oak]

There is no such thing as that. Where are you trying to download it from

[drowsy scaffold]

w-okoda

[rare goblet]

Ayo? @drowsy scaffold level 1 !!!

[long matrix]

Thanks for the help

[rare goblet]

Ayo? @long matrix level 1 !!!

[drowsy scaffold]

could u jus show me a whole tut on how to get the software

maybe screenshare idm

[knotty moth]

please read the guide thoroughly before asking nonsense

[pastel oak]

Did you open the guide I linked you

[drowsy scaffold]

whatt

wdym

i found out abt w-okoda by some ytber

[pastel oak]

Oh, I linked it to Mitch, not you, I got you mixed up
https://rentry.co/ForkVoiceChangerGuide

Dont rely on youtube videos they are outdated

[drowsy scaffold]

alirght ty

[pastel oak]

We will do our own youtube tutorials soon hopefully xd

[drowsy scaffold]

wait

so how exactly do i set up the whole thing??

the WHOLE thing

im lost 😭

[long matrix]

You got any idea how I can make the program output to an audio device? As I can't find one its generating if that makes sense

[pastel oak]

Navigate to "Download" and get the ones for your GPU
Navigate to "Virtual Audio Cable" if you do not have that yet
Navigate to "Opening" to know how to extract the file, and which file to run

Then upload voice model if you dont know how to, then Audio Setup for the inputs and outputs

[drowsy scaffold]

right ill try that

[pastel oak]

I dont understand exactly. Do you mean how to output the voice changer to discord/games for example?

[long matrix]

Yes

[pastel oak]

Get the Virtual Cable from the guide, then do the first thing it says on Audio Setup

[agile gale]

what does audio blocked mean

[pastel oak]

Are you using RVC GUI Realtime?

[low shard]

yea no that's not good enough for local training, you will have to do it on cloud

[agile gale]

yes

[drowsy scaffold]

@pastel oak

[agile gale]

just says io block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.
Audio block passed.

[drowsy scaffold]

should dis be hapening

[pastel oak]

Thats no error, its loading up afaik havent used in a while

[low shard]

As you dont got a good PC, its better you use cloud for training an RVC Voice Model:

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
• Applio (ui)
• Mainline (UI)
• RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
• Mainline (UI)

[pastel oak]

Yes its downloading files

[drowsy scaffold]

ok

[agile gale]

Oh so thats normal its just loading?

[rare goblet]

Ayo? @agile gale level 1 !!!

[pastel oak]

goes straight into my copy paste list i love you

[pastel oak]

It should follow up with new information right after that

[low shard]

[pastel oak]

Anything new on the console yet after audio block passed

[agile gale]

how lonjg should it take?

[pastel oak]

Few seconds

[agile gale]

its just spamming audio passed

nothing else

[pastel oak]

Can you send a screenshot of your GUI, and of the console

[agile gale]

idk which to use

[pastel oak]

Use (MME)

On both

I notice your RVC is saved in OneDrive, I am not sure if that actually works, but try with (MME) and then see if it starts to say anything else after few seconds of audio block passing

[agile gale]

@pastel oak

[steep mantle]

I want to know how to make pth and index file

[pastel oak]

You need to change output to MME aswell, like I said. You only changed one of them, this causes it to crash

[pastel oak]

Whats your GPU

[drowsy scaffold]

'shad what to do from here im lost again

[agile gale]

i dont think ill hear anything if i do that lol

[steep mantle]

GTX

[pastel oak]

You should.

[agile gale]

i cant

lol

[drowsy scaffold]

still lost could u screenshare or smth

[pastel oak]

Its better you use cloud for training an RVC Voice Model instead of local training:

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
• Applio (ui)
• Mainline (UI)
• RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
• Mainline (UI)

[agile gale]

when i changhe to

i cant hear anything

[knotty moth]

yeah just rename the extension to .pth
so please read the guide thoroughly

[pastel oak]

Im out of country, dont have a setup

Literally click on select file

💀

[primal fulcrum]

I did but still found it frustrating

[knotty moth]

skill issue

[agile gale]

@pastel oak cant hear

[pastel oak]

both input and output have to be of same type, MME is the safest. I dont understand what you mean by this, have you tried selecting input: microphone (MME) and output: speakers (MME)? Scroll up to the top if needed

[agile gale]

i cant hear anything

when i put my output to this

where should i be bluhhed my headphones

plgging

cant hear non lool

[pastel oak]

Is that even your headphones

the DP thing that doesnt look right

[agile gale]

thimk thats my monitor

[pastel oak]

Do you have headphones plugged in? If yes, make sure to select those as output then

Technically you will need to select your virtual cable as output to be using this on discord and games, but for now headphones to test

[agile gale]

now i9 cant hear

@pastel oak hiw kllong this gonna go for?

[pastel oak]

This is a sign that it works

Whats your GPU?

[agile gale]

4060

[pastel oak]

sample length increase it to 0.15

[drowsy scaffold]

ok @shad now that iv got it working and i know what it sounds like what should i do now

[agile gale]

@pastel oak I HEAR HELLA LIKE

[rare goblet]

Ayo? @agile gale level 2 !!!

[pastel oak]

Send screenshot of how your voice changer looks like rn

[agile gale]

tikking

sounds

noises

[pastel oak]

Did you increase the sample length

[agile gale]

yup

[pastel oak]

Select your GPU

[agile gale]

i hear ltos of noises tho

[pastel oak]

Your microphone is picking up a lot of sounds

[drowsy scaffold]

yup what now

[pastel oak]

Since you selected the output to be your monitors speakers, it is also picking up the converted voice back again in a loop

[pastel oak]

Which gpu do you have

[drowsy scaffold]

intel uhd grap 620

[rare goblet]

Ayo? @drowsy scaffold level 2 !!!

[drowsy scaffold]

💀

[pastel oak]

damn

[drowsy scaffold]

ik its bad

[pastel oak]

intel uhd dont run well at all but lets see what we can do

I tell you what to do

[drowsy scaffold]

bruh

when i run gpu i hear no voice being changed

[pastel oak]

Chunk: 1000 ms
Extra: 0.5s
Pitch: 12 (youre using a female model)

Open "advanced settings" on the bottom
Click on "convert to onnx", this will take 1 minute to load up

Then try the voice changer, and tell me what number the "perf" value on the top left of the picture shows, and what color

[knotty moth]

I bet it wont be better than cpu mode

[drowsy scaffold]

oh

[agile gale]

@pastel oak how do i know the right pitch for all models

[pastel oak]

Actually could be true. I believe deiteris mentioned that Intel UHD Graphics dont run good at all, this is an issue he might adress in the future

[agile gale]

dont they say it?

[pastel oak]

its dependant of your voice and how high or low pitch the voice model is. So you have to trial and error until it fits

[agile gale]

how do i get it to work on liek discord

[pastel oak]

https://rentry.co/RVCRealtimeGuide

go to virtual cable, install that

then do this

[agile gale]

@pastel oak ?

this correcft

[drowsy scaffold]

erm its just stuck on this i think

nvm

just went

shad i cant hear my voice bruh

[pastel oak]

yes

[pastel oak]

Idk man intel uhd cant run this i think xd

but connection error means you maybe closed the console window

is it still running?

[agile gale]

@pastel oak like this

[drowsy scaffold]

yeah

[pastel oak]

Correct

[pastel oak]

gg

Dont close it xd restart it

[pastel oak]

and on rvc gui you select Line 1 (mme) on the Output instead

[drowsy scaffold]

ow do i do that

[pastel oak]

The same way you started it

[drowsy scaffold]

aint nothing happening bruh

[pastel oak]

Show perf number on top left

[pastel oak]

did you click start....

green button

[drowsy scaffold]

yupp

[pastel oak]

change the gpu to cpu

try with that

[drowsy scaffold]

thats what i was thinking

i started hearing nothing when i switched t gpu

but the operating in progress thing will take a while

[wispy lodge]

Could you send vcclient.log that is in the voice changer folder?

[pastel oak]

Ill let the developer know if its an intel uhd specific problem or not

Ah there he is

[drowsy scaffold]

ok

IT WORKS

[rare goblet]

Ayo? @drowsy scaffold level 3 !!!

[drowsy scaffold]

@pastel oak how can i make it work for discord,roblox etc

[pastel oak]

Increase the chunk value to 1700, it should make the perf number green

[pastel oak]

Did you download virtual cable from the guide

[drowsy scaffold]

lemme see

[knotty moth]

it will be crysis meme if you try running for roblox

[pastel oak]

I have to warn you, running cpu for voice changer rlly is not a great idea

[drowsy scaffold]

wdym

[pastel oak]

especially if you try to run a game

[drowsy scaffold]

how

what will happen 😭

[pastel oak]

explosion in a 2 block radius

[drowsy scaffold]

bro fr?

nah be fr

[long matrix]

Is there a way to use integrated graphics to process the voices?

[pastel oak]

no lol but like youre running an AI software on a cpu, its not made for it

[drowsy scaffold]

maybe it is for inteluhd???

[long matrix]

Nah im running it on a gpu

[knotty moth]

if it is at least radeon 780M iirc

[long matrix]

I have intel uhd 770

[pastel oak]

nah definitely not for intel uhd either lmao u need a dedicated graphics card like Nvidia RTX or AMD RX

[drowsy scaffold]

could u just help me use roblox with it

[long matrix]

my cpu is the i7-14900k

700k*

[pastel oak]

download virtual cable, then do what it says on "Audio Setup" first step

[drowsy scaffold]

ok

[long matrix]

How

[pastel oak]

Youre a maniac for using a cpu like that with an integraded card

You might aswell use CPU conversion

[long matrix]

I have a 4070 but I wanna play using that

I lag while using the software

[pastel oak]

You should 100% use the 4070, you can run the game with it aswell

[drowsy scaffold]

waitwhere do i download virtual cable

[long matrix]

Idk lags like hell when I use the cpu conversion

[pastel oak]

Send screenshot of your settings

while using nvidia one

[pastel oak]

on the guide

[knotty moth]

you can use your 4070 for most non-demanding games (CS, valorant, etc)

[drowsy scaffold]

this??

[long matrix]

Ok thanks

[rare goblet]

Ayo? @long matrix level 2 !!!

[pastel oak]

Brother thats what u already downloaded lmfao go to virtual audio cable

i cant babysit you through this

[drowsy scaffold]

this??

[long matrix]

My integrated is doing nothing

[long matrix]

Thanks so I can't use my uhd 770 integrated for this?

[pastel oak]

Lets start from the beginning

• you should use the 4070, even running a game next to it is more than fine. intel uhd does not work + its the worst thing you can choose

• if you lag, you may reduce ingame quality or limit fps, and raise chunk if needed

[pastel oak]

ur trolling me, how old are you

[knotty moth]

I see no point of utilizing it other than some basic display, streaming

[drowsy scaffold]

lol yh i am

fine so what do i do now

[long matrix]

Alright if I plug another gpu in as I have a base gtx 970 laying around somewhere can I use that solely for the voice?

[drowsy scaffold]

bro shad r u there

[knotty moth]

that's too old unfortunately, again why do you avoid using 4070? what game are you playing?

[long matrix]

Red dead redemption 2

Quite demanding on my gpu is all

[knotty moth]

try lower some settings and fps cap to 60

[long matrix]

Ok I'll do that now

Is there a way to make the voice changer run more efficiently or use less power

[pastel oak]

@drowsy scaffold

[drowsy scaffold]

yh

do i run setup.exe

[pastel oak]

You can raise chunk (increases delay), or reduce extra (reduces quality)

[pastel oak]

it says what to run on the guide man

[drowsy scaffold]

ok

[pastel oak]

Tf do i write guides for

😭

[knotty moth]

I think 4070 will barely have temp issue

[long matrix]

It's not the temp as I have a good AIO it's more the fps I get is like 20 when in game

[lethal orbit]

hello its the same issue with my friend

i tried everything

but his voice keeps on cutting out

[knotty moth]

ah.. you'd be more likely to need a decent 2nd RTX laptop/pc

[pastel oak]

f0 det: rmvpe

[pastel oak]

gtx 970 works on this version of wokada, no clue about benchmarks

[drowsy scaffold]

works like butter shad thanks

😭

[long matrix]

Okay I might plug it into a spare PCIE and try it. Thanks.

[drowsy scaffold]

hope i dont crash in rblx or smth

[pastel oak]

Nice, good luck

[regal axle]

didnt work

[rare goblet]

Ayo? @regal axle level 2 !!!

[regal axle]

still is cutting out

[pastel oak]

What game are you running next to it

Disable "echo" and use headphones if you arent

[regal axle]

im using headphones

[pastel oak]

Record audio of how it sounds like cutting out

[regal axle]

no game running in the background

[pastel oak]

Increase chunk to 112

[knotty moth]

lower threshold and increase chunk

[regal axle]

wait there is no cutting out when i record it but in discord its cutting out?

[drowsy scaffold]

wait last thing

should i enabke passthru

or nah

[pastel oak]

No

[pastel oak]

Turn off Echo Cancellation on discord
Turn off noise suppression on discord

Move the sensitivity slider to the left on discord

[regal axle]

ok did that lemme check

[wise roost]

I have a question, when I use pretrained models like Titan or Ov2Super, the inference output generated using the models trained using these pretrained sounds more robotic (and similar) in comparison with using the original RVC2 pretrained. Is it expected or am I doing something wrong here?

[pastel oak]

Original pretrain is the best pretrain

Custom pretrains need better cleaning of dataset for sure

[regal axle]

i did all that didnt work :/

im on 128 i was on 96 how many chunks should i use

[pastel oak]

Can your friend record you

Are you sure its cutting out and not crackles

[wise roost]

ah! ok. I thought using these pretrained would do a better job than the original ones.

[rare goblet]

Ayo? @wise roost level 1 !!!

[wise roost]

I was trying to explore if there's a way to train the model using less audio input than 7 mins without compromising too much on the quality.

[pastel oak]

Yeaa hifigan is crap unfort

[rare goblet]

Ayo? @lethal orbit level 2 !!!

[lethal orbit]

heres the recording

[pastel oak]

I mean my titan trained models on 5 mins come out good on inferencing so idk

[lethal orbit]

look at de recording

[pastel oak]

Thats crackles

Open Task Manager > Details

Right click audiodg.exe and set priority to High

Right click audiodg.exe again > set affinity > uncheck everything except CPU 2 (only ✅ CPU 2, turn off the rest)

With a program called ProcessLasso you can automate this to always be active, since Windows resets these sometimes.

[wise roost]

oh! interesting. Do these pretrained models require any change in the inference parameters in comparison with original rvc?

Because for me all the outputs sounded very similar and no where closer to the original voice.

[pastel oak]

I am not quite sure, I usually use default except I turn off index usage and I match the volume to the original inferenced audio

I wouldnt think they need different changes in parameters necessarily. I turn off index because I dont like the original accent of voice models for my usage, it can reduce artefacts though

[wise roost]

ok, how to turn off index usage? And what do you mean by that?

[pastel oak]

the default is 0.75 iirc, i put that to 0 to disable it

And the volume thing, I dont remember the name off the top of my head, you can send a screenshot of the parameters I should be able to identify it

Im out of country, cant check myself

[wise roost]

ok, let me share

[pastel oak]

Those are training parameters, I thought we were talking about inferencing

[wise roost]

ah! I'm sorry. I was talking about training the model and then generating inference using the pth and index files.

[pastel oak]

And you were saying you trained a model on original pretrain, Ov2 and titan ; and in the inference process, ov2 and titan did not come out as good as on pretrain, correct?

[wise roost]

that's correct!

[pastel oak]

The index and volume thing I described are inferencing parameters

However, I am noticing this is not making much sense

xd

[wise roost]

I think maybe I don't know what I'm doing then.

[pastel oak]

I think you should stick to original pretrain, since it is the best

Maybe someone else, who is far far more experienced than myself can comment on why ov2 and titan on inference did not come out as greatly, aside from them not being as good as original

like atri 🙏

[knotty moth]

there's no relationship between inference parameters and pretrains
also it mostly depends on the dataset quality and variation

[wise roost]

got it. Considering the dataset quality as a constant, the inference generated using the models trained using original pretrained is coming out to be better than titan and OV2.

[knotty moth]

they might have some pros and cons for average and very clean datasets as well, though anyway original pretrain is still good to go if you're not sure what you're doing

[rare goblet]

Ayo? @knotty moth level 29 !!!

[serene hound]

it was working fine few weeks before

[pastel isle]

Hello can someone help me getting RVC 2 to run on a Steam Deck please?
https://github.com/Mangio621/Mangio-RVC-Fork/releases

[knotty moth]

never
at least inference is still doable in cpu mode

[dire cipher]

hi smart ai people, im trying to install rvc rn on my new laptop but it's not willing to recognise that requirements.txt exists.

[tame mica]

omg

[dire cipher]

OMG IT'S YOU HELLO

i did not expect to see you here

[dire cipher]

so real

[rare goblet]

Ayo? @dire cipher level 2 !!!

[dire cipher]

GO AWAY IRRELEVANT BOT

no but yall, it's trying to be funny like that

[pastel isle]

Yh thats what i mean inference version.

[serene hound]

There is a easier way to download update without need to download entrie applio again?

[odd shale]

I don't think so.

[serene hound]

oh..

[azure marsh]

Command has expired

[hoary gyro]

nevermind found the answer in guides

actually there is no answer

how do you convert a voice present day cause last time i used it was with google colab with rvc 2.0 so im completely lost

[flint geyser]

heya, does anyone need help?

[flint geyser]

check out https://aihub.wtf/docs

[hoary gyro]

thank you

[empty fjord]

I want to know why the data related to RVC, which should have appeared in the top left corner, isn't showing up on mine, so no matter how I adjust it, I can't hear any sound

[hushed sedge]

hi

is 20 minutes long enough for a dataset?

[primal fulcrum]

Where I get Index data?

[carmine hearth]

Is there a way to use harvest in Applio?

[low shard]

i don't know which one of the cloud ways are you using but when training you there would be another button to train the .index

[odd shale]

Yep, it is. Make sure it's free of any kind of noise tho + apply audio labeling to the audio with Audacity and export it to a folder.

[pastel oak]

Can you send a full screenshot of the voice changer? Which version did you download?

[odd shale]

You must hit the "train index" button before training the model.

[hushed sedge]

does someone know what this warning mean?

C:\Users\user\Downloads\Applio-3.2.0\env\lib\site-packages\torch\autograd_init_.py:251: UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance.
grad.sizes() = [64, 1, 4], strides() = [4, 1, 1]
bucket_view.sizes() = [64, 1, 4], strides() = [4, 4, 1] (Triggered internally at C:\actions-runner_work\pytorch\pytorch\builder\windows\pytorch\torch\csrc\distributed\c10d\reducer.cpp:334.)
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass

[pastel isle]

I got Applio running on the Steam Deck but gives me an error when trying to convert something.

[tidal trail]

xd

[tidal trail]

that is not an error

[fossil vigil]

hey can i get some help with rvc

oh wait what

i installed python and all

and no app is showing

just command prompt

theres an installation there and im waiting for it to finish

[rare goblet]

Ayo? @fossil vigil level 3 !!!

[fossil vigil]

could be it

[pastel oak]

Wait for it to finish

[fossil vigil]

oh ok

im following this tutorial
https://www.youtube.com/watch?v=We5oYpCR3WQ&list=WL&index=4&t=2s

[pastel oak]

tutorials bad

https://rentry.co/VoiceChangerGuide

[fossil vigil]

oh yeah i have a amd gpu and installed cuda

it worked fine

idk how

well

it worked lol

so uh

[empty fjord]

The version I'm using is: vcclient_win_std_2.0.58-alpha.

[fossil vigil]

im testing the demo it isnt working

[pastel oak]

You didnt select a GPU

And the alpha version is unstable

[pastel oak]

whats your gpu name

[fossil vigil]

rx 580

8 gb

[pastel oak]

f0 det: rmvpe_onnx
chunk: 256
extra: 8192

and make sure u selected the right gpu number

[fossil vigil]

it should be the gpu number in task manager right?

[pastel oak]

yes

original wokada sucks for amd though youd get much better performance on the modified if you wanna download that

[fossil vigil]

sure

i tried doing the voice changer

it wont open

i meant like no sound

[pastel oak]

https://rentry.co/ForkVoiceChangerGuide

[fossil vigil]

im talking

but theres nothing

[pastel oak]

get the other version i just linked anyway

[pastel oak]

the no sound can be fixed later we can look at it

[fossil vigil]

okay

do i close down this?

[fossil vigil]

this download?

[pastel oak]

yes

[pastel oak]

yes

[fossil vigil]

ok itll be downloaded in like a couple seconds

[fossil vigil]

what now

do i extract it?

[pastel oak]

Yes

then run mmvcserversio.exe

[empty fjord]

I did select a GPU, but it still doesn't work. Also, my friend downloaded this version, and it works for them

[pastel oak]

then read the guide on uploading voice models & the first part of audio setup

[pastel oak]

Yh the alpha is unstable as i said, for some it works for some it doesnt

[fossil vigil]

uploading voice models shouldbe easy did it before

[pastel oak]

best of using this
https://rentry.co/ForkVoiceChangerGuide

[fossil vigil]

this happened when opening it

pressed enter and it just closed

[pastel oak]

Did you have an antivirus program running that may have deleted files?

[fossil vigil]

i replaced the previous files that i had

but no

just windows antivirus

which does nothing

[wispy lodge]

Don't replace. Just unpack into a separate folder

[pastel oak]

I was about to ping the developer, but he is already typing 🙏

[fossil vigil]

o ok tysm

this thing right?

[rare goblet]

Ayo? @fossil vigil level 4 !!!

[wispy lodge]

Yeah

[fossil vigil]

okay its installing the thing

[wispy lodge]

Noice

[hard bloom]

help, in the voice changer program only 1 standard voice works, all the others do not work. I have already installed new ones and there is 0 sense, maybe someone had such a problem?
in the command line it says "Pipeline is not initialized" and so on endlessly

[pastel oak]

Install failed, delete pretrain and model_dir folders and then launch start_http again

[fossil vigil]

should i just wait

oi\

it opened google

am i supposed to use that

this opened

[empty fjord]

Sure, thanks! I'll give it a try.

[wispy lodge]

Yes, it uses WebUI that runs in the browser

[fossil vigil]

woah cool

gonna download some voice models and ill be back

[hard bloom]

thank you very much, you are the best thing that happened today :3

[fossil vigil]

tysm it worked

this is fun

how do i use it on discord

if i wanted to

or in vc

[empty fjord]

How do I download this? I couldn't find where to download the Voice Changer software

[wispy lodge]

You need a virtual cable installed like VAC. Then set output to line 1 in voice changer, and input to line 1 on discord/other apps

[fossil vigil]

oo

i have virtual cable but never really opened it

[pastel oak]

Open the guide i linked, go to download section etc

[pastel oak]

You dont open a program regarding vb cable, just use it in the app and games/discord

[empty fjord]

I only see options for downloading NVIDIA, AMD, INTEL, and CPU, and Mac. I don't see a download for Windows.

[pastel oak]

They are all windows except mac

These are hardware specific downloads, as they run on GPU

[violet heron]

You download depending on your GPU

[pastel oak]

Open task manager > performance

check for your gpu0 and gpu1 (if it exists)

[violet heron]

Nvidia, AMD, Intel are GPU brands

[empty fjord]

thanks! I'll give it a try

[rare goblet]

Ayo? @empty fjord level 2 !!!

[fiery raptor]

the refresh button on my applio (the colab version) is not working