✨│ai-help

[verbal ridge]

Can I get some help getting Applio to open? Solved: open the applio manager batch file and select update

2023-10-08 17:57:20 | INFO | assets.configs.config | Found GPU NVIDIA GeForce GTX 1060 6GB, force to fp32
Traceback (most recent call last):
  File "S:\Applio-RVC\Applio-RVC-Fork\infer-web.py", line 55, in <module>
    from lib.infer.modules.vc.modules import VC
  File "S:\Applio-RVC\Applio-RVC-Fork\lib\infer\modules\vc\modules.py", line 22, in <module>
    import tabs.merge as merge
  File "S:\Applio-RVC\Applio-RVC-Fork\tabs\merge.py", line 15, in <module>
    import tabs.resources as resources
  File "S:\Applio-RVC\Applio-RVC-Fork\tabs\resources.py", line 49, in <module>
    config = Config()
  File "S:\Applio-RVC\Applio-RVC-Fork\assets\configs\config.py", line 82, in wrapper
    wrapper.instance = func(*args, **kwargs)
  File "S:\Applio-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 "S:\Applio-RVC\Applio-RVC-Fork\assets\configs\config.py", line 235, in device_config
    self.use_fp32_config()
  File "S:\Applio-RVC\Applio-RVC-Fork\assets\configs\config.py", line 202, in use_fp32_config
    with open("infer/modules/train/preprocess.py", "r") as f:
FileNotFoundError: [Errno 2] No such file or directory: 'infer/modules/train/preprocess.py'```

[wheat copper]

is it ok for my samples to have parts with no audio?
or do i have to edit it all

[proud elbow]

do truncate silence using Audacity

[shut axle]

still want an answer to this!

[rare goblet]

@shut axle has reached level 1.

[wheat copper]

how?

[proper shale]

Effects > Truncate Silence

[rare goblet]

@proper shale has reached level 19.

[wheat copper]

thanks!

[brittle wing]

I set things up etc but when I press start it doesnt change my voice at all what I can do

[mossy pasture]

you are missing a preprocess.py file?

[verbal ridge]

Apparently, but there's no folder named infer in the top directory

[mossy pasture]

then you probably missed the whole folder

go to your applio folder, then find lib\infer\modules\train, and see if preprocess.py exist

if you don't have it, you can find it on the applio rvc fork github and download it manually

[verbal ridge]

Someone in another server helped me out and it's working now

[rare goblet]

@verbal ridge has reached level 6.

[wheat copper]

what does this mean?

RuntimeError: The expanded size of the tensor (17280) must match the existing size (0) at non-singleton dimension 1. Target sizes: [1, 17280]. Tensor sizes: [0]

[loud hazel]

u use old *.pt, get new one

[wheat copper]

i already got it working thx tho

[rare goblet]

@wheat copper has reached level 4.

[rugged stirrup]

Can someone tell me whats wrong with this?

[cyan spear]

how do i solve this?

[violet heron]

connect gpu

top right

[austere stream]

question

how much ram do you need to run locally?

[wide parcel]

so i made an rvc model and tried to submit it since it sounded fine (i understand its overtrained) but i tried using it in realtime voice changer and a diff webui and it sounds horrible
but it sounded fine when i inferred on the one i trained it on

any reason for that?

[cyan spear]

wouldnt let me

all the gpu options were blacked out

[violet heron]

use alt account

[cyan spear]

ok

is it this?

[mortal matrix]

yo i be getting TONS of edlay while playing games with the voice changer

please someone tell me how to fix

[cyan spear]

do you have a powerful pc?

[mortal matrix]

yes

[cyan spear]

what graphics card n shit

[mortal matrix]

1080

[cyan spear]

idk

i got a 3060 and mine is js fine

you just need a super powerful computer

[mortal matrix]

im offing myself

[rare goblet]

@mortal matrix has reached level 1.

[cyan spear]

rest in peace fr

[mortal matrix]

[cyan spear]

[rugged shuttle]

@prime hemlock how do i make a ai voice model without having to work on the dataset for more than 3 days

[prime hemlock]

https://tenor.com/view/tommy-boy-chris-farley-no-idea-clueless-gif-4280737

[rugged shuttle]

meanie

[rugged stirrup]

Could someone help with this

[proper gust]

i get this error when trying to install pytorch ERROR: Could not find a version that satisfies the requirement torch (from versions: none)
ERROR: No matching distribution found for torch

yall know a fix?

[slow summit]

whats the diff between 840 and 980 in a PTH File?

[bold ermine]

Guys I have implemented notebook on azure, can anyone assist me where I have to put pretrained model in, azure is not using collab it is using own space so which folder I have to place pretrained model so I can have in model inference tab

[plush tree]

Hello , when I download an rvc on aicovergen, it takes too long to download, can anyone help me?

[rare goblet]

@plush tree has reached level 1.

[frigid wasp]

Is anyone else having issues with RVC on HuggingFace?

[plush tree]

Yes

[frigid wasp]

It's incredibly slow on my end

[rare goblet]

@frigid wasp has reached level 1.

[plush tree]

Mine was taking too long to download

An rvc model

So i switch up to a new browser

[frigid wasp]

It takes forever to make AI covers

[terse nymph]

yeah all the links always take forever for me i just gave up

[plush tree]

Me neither, it takes forever

Sometimes i get an error

[serene musk]

3 days ?? Golllyy

What are you doing to make it take 3 days to make a dataset?

[torn valley]

im on the RVC hugging face thing and when I insert my download link, it says "There's been an error."

[glad zealot]

which one

[torn valley]

its a peter griffin one, i can download the file though

in the huggingface logs its saying error 403: forbidden

so maybe it doesnt have some kind of perms to grab it?

[bold ermine]

Guys I have implemented notebook on azure, can anyone assist me where I have to put pretrained model in, azure is not using collab it is using own space so which folder I have to place pretrained model so I can have in model inference tab

[torn valley]

i set it to public and restarted it, any fix?

[rare goblet]

@torn valley has reached level 1.

[glad zealot]

????

[rugged stirrup]

Does anyone

[serene musk]

I'm so curious what's going on here lol

My curiosity is insatiable

[woven atlas]

how i can get a rvc maker rol

?

[marble timber]

guys how to get model master role?

[serene musk]

Post a model in the get model maker tab

https://discord.com/channels/1159260121998827560/1159514067187277865

[rugged stirrup]

Does anyone know how to setup a public URL on RVC

[sand iris]

as in ? Running RVC locally and making it available to others ?

[rugged stirrup]

to others

[rare goblet]

@rugged stirrup has reached level 4.

[rugged stirrup]

Oh and also

[sand iris]

look into ngrok, it would allow you to "open" a port without actually doing port forwarding

[rugged stirrup]

If you could take a look at this whats wrong in that

[rugged stirrup]

in the mangio rvc folder?

[sand iris]

It's an external software

[rugged stirrup]

Oh I dont have that then

[burnt agate]

wouldnt a public gradio link be easier tho?

[sand iris]

Hang on a sec, just gotta look if there isn't an easier way

yeah

gradio public URLs are a thing

[rugged stirrup]

Cant find a thing called share()

[burnt agate]

just add --share behind the command like so python infer-web.py --share

just a example ^

[rugged stirrup]

wait just click the .bat?

it doesnt let me input anything

[burnt agate]

could edit the bat file and add it in

[rugged stirrup]

the .bat is 1kb

[sand iris]

the bat only contains this :

runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
pause

Just edit it to this :

runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897 --share
pause

[rugged stirrup]

Alright ill see if that works

[burnt agate]

mine doesnt even look like that anymore lmao

[rugged stirrup]

I pressed a key just closed

[sand iris]

Of course it has to be a bit harder lol

[rugged stirrup]

I just pasted it in notepad and saved it

should work right?

[burnt agate]

why does that not work

[sand iris]

There's no share flag lmfao

jank

[sand iris]

https://github.com/Mangio621/Mangio-RVC-Fork/blob/main/infer-web.py#L3083

Look at this

https://cdn.discordapp.com/emojis/790214199544512573.webp?size=48&name=kek&quality=lossless

[burnt agate]

booo

[sand iris]

Okay time for some dirty editing ig

[rugged stirrup]

I am Not smart with code

[burnt agate]

I swear there was a way I did it once

[sand iris]

It's gonna be fine

can you open the infer-web.py file ?

[rugged stirrup]

yeah

what do you want me to look for in it

[sand iris]

Now scroll down to the end of it

[sand iris]

    if (
        config.iscolab or config.paperspace
    ):  # Share gradio link for colab and paperspace (FORK FEATURE)
        app.queue(concurrency_count=511, max_size=1022).launch(share=True)
    else:
        app.queue(concurrency_count=511, max_size=1022).launch(
            server_name="0.0.0.0",
            inbrowser=not config.noautoopen,
            server_port=config.listen_port,
            quiet=False,
        )

should be the last few lines

yeah

    if (
        config.iscolab or config.paperspace
    ):  # Share gradio link for colab and paperspace (FORK FEATURE)
        app.queue(concurrency_count=511, max_size=1022).launch(share=True)
    else:
        # app.queue(concurrency_count=511, max_size=1022).launch(
        #    server_name="0.0.0.0",
        #    inbrowser=not config.noautoopen,
        #    server_port=config.listen_port,
        #    quiet=False,
        #)
        app.queue(concurrency_count=511, max_size=1022).launch(share=True)

[rugged stirrup]

Just paste it there?

In this area

[sand iris]

replace the last block with what I sent

[rugged stirrup]

Alright

[sand iris]

Now it should work

also undo the --share trick in go-web.bat

[rugged stirrup]

Hold up lits doing something

[burnt agate]

ahh I just realized, im thinking of aicovergen that works like that without editing anything

[rugged stirrup]

Whats it doing

Oh wait

Thanks

it works

[sand iris]

Same thing as it always did, but now it has a public url

yay

[rugged stirrup]

Why is it always people with CDI pfps that help me

So does that mean I can give someone this link and let someone train or inference off my computer?

[burnt agate]

yup

[rugged stirrup]

Awesome

I also get a strange error when i try to use a more than 30 minute dataset

How could you let someone upload a song without having to have it on a computer

[sand iris]

I think applio has the feature to make the server download

[rugged stirrup]

Applio is impossible to use

[sand iris]

https://cdn.discordapp.com/emojis/692987693944012811.webp?size=48&name=KEKW&quality=lossless

[lost quartz]

you should probably learn machine learning before saying this..

[rare goblet]

@brittle wing has reached level 1.

[lost quartz]

it is a bit complicated, sometimes more number of epochs can lead to more quality, sometimes not

so far 600 epochs should be okay

[bright venture]

is AMD supported on mangio?

amazing info, please marry me

yes

[marble timber]

that is model maker, and im curious how to get model master

[serene musk]

Make alot of popular models

[uncut pasture]

Does it matter when you "Train Feature Index", like, before doing the actual model training or after?

[potent root]

doesn't actually matter in terms of quality

but i prefer doing it before

[uncut pasture]

This was my assumption, I have been doing it before anyway

[rare goblet]

@uncut pasture has reached level 1.

[uncut pasture]

thanks

[potent root]

training index before training the pths gives you the option of testing your model

[rugged stirrup]

Could somebody help me with this

[loud berry]

Cristiano Ronaldo model please?

[rugged stirrup]

this isn't model requests

[loud berry]

where to get?

[rugged stirrup]

model requests

or just look in the model channel The voice model channel you'll probably be able to find it

[surreal bluff]

from where to get these hugging face voices for api

[surreal bluff]

where is the moddle channel

[grand coral]

I've got 77 minutes of fairly clean audio, what epoch amount would you guys recommend and is 77 minutes too much?

[fossil glen]

a

[potent root]

large datasets, at some point, have diminishing returns

i think 77 minutes is doable

[grand coral]

so should I bump it down to 45 or 50

[potent root]

but probably not worth the time it takes to train imo

it's up to you to decide

if you're gonna go with 77, i'd suggest 500 epochs

[grand coral]

I think 45 is more reasonable in this case and all I have to do is just trim down the recording, don't want to risk it

[potent root]

yeah

[grand coral]

so for 45 Im gonna take a guess for around 350ish epochs would probably be a good option

[rare goblet]

@grand coral has reached level 1.

[grand coral]

if 500 is recommended for 77 then it should be fine using around 350 for 45 considering how I've seen some people use 300 for 10 minutes of audio

[potent root]

do more epochs if it's a shorter dataset!

just let it be known that you have to consult the graph and not rely on the final epoch version of the model

[grand coral]

alright

[proper shale]

+, this number isn't definitive

can vary a lot with batch_size and shit

[dim rain]

are there any working colab for training

[woven atlas]

can someone send link to rvc disconccted train

?

[wispy burrow]

#1159949278270193734

[trail oriole]

applio (running on runtime) errors - inference
there's not really much info i can give yall besides these errors

2023-10-09 21:34:48 | WARNING | lib.infer.modules.vc.modules | Traceback (most recent call last):
  File "D:\Applio\Applio-RVC-Fork\lib\infer\modules\vc\modules.py", line 375, in vc_single
    audio_opt = self.pipeline.pipeline(
  File "D:\Applio\Applio-RVC-Fork\lib\infer\modules\vc\pipeline.py", line 637, in pipeline
    and os.path.exists(file_index)
  File "genericpath.py", line 19, in exists
TypeError: stat: path should be string, bytes, os.PathLike or integer, not list

Traceback (most recent call last):
  File "D:\Applio\Applio-RVC-Fork\runtime\lib\site-packages\gradio\routes.py", line 437, in run_predict
    output = await app.get_blocks().process_api(
  File "D:\Applio\Applio-RVC-Fork\runtime\lib\site-packages\gradio\blocks.py", line 1349, in process_api
    data = self.postprocess_data(fn_index, result["prediction"], state)
  File "D:\Applio\Applio-RVC-Fork\runtime\lib\site-packages\gradio\blocks.py", line 1283, in postprocess_data
    prediction_value = block.postprocess(prediction_value)
  File "D:\Applio\Applio-RVC-Fork\runtime\lib\site-packages\gradio\components.py", line 2586, in postprocess
    file_path = self.audio_to_temp_file(
  File "D:\Applio\Applio-RVC-Fork\runtime\lib\site-packages\gradio\components.py", line 360, in audio_to_temp_file
    temp_dir = Path(dir) / self.hash_bytes(data.tobytes())
AttributeError: 'NoneType' object has no attribute 'tobytes'

[wooden pine]

I am currently re-installing the whole RVC program because when I get prompted the GUI for using the realtime AI Voice overlay it just says "pause" and stops working

any fix?

[autumn hill]

dunno if this is the proper place but what do i do if i want my model master back but i haven't uploaded my models i had from the old server

[short ridge]

-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

[mellow peak]

I can't use the RVC collab for covers , only training models?

[rare goblet]

@mellow peak has reached level 1.

[mellow peak]

don't know if I'm saying it correctly

Because I was using another collab for RVC that I used days ago to make covers but I can't get past this section, it just stops there, help and guidance would be appreciated thanks ToT

I'm considering to install RVC locally If I can't use the collab:c

[proper shale]

Working colabs are on #1159949278270193734

RVC Disconnected is for training specifically

[spice pagoda]

What epoch is recommended for short and long datasets? I'm kinda confused, since my new model results is really bad in quality. Just send me a list of recommended epochs for short and long datasets to make it better.

Here's my new model and listen to the demo I uploaded: https://discord.com/channels/1159260121998827560/1160962701179224196

[rare goblet]

@spice pagoda has reached level 2.

[mossy pasture]

epoch numbers are not important at all, it's more important that you set a higher save frequency and also use a better quality dataset

with higher saving frequency, you can have more models, and search for the best one

[proper shale]

There isn't a set number for both of those - they vary

[silk ferry]

I literally remember seeing a jameskii ai model why isn't it here

[proper shale]

Some models weren't backed up

[silk ferry]

Mann I really wanted that model ngl

[mossy pasture]

you can make it yourself :)

[silk ferry]

I donr have a pc rn 😭

[mossy pasture]

colab is another option

[silk ferry]

Colab?

[proper shale]

if they dont have a pc they wont be able to process the dataset stuff correctly

[mossy pasture]

is it? cus it only uses the google drive

[rare goblet]

@mossy pasture has reached level 3.

[mossy pasture]

and i think you can use 7zip or any apps to manage the phone folder system

[proper shale]

but audacity is what they need basically

[mossy pasture]

oh yea, i forgot about that part

[silk oasis]

how to make ai covers when collab got cancelled and applio doesnt work properly?

[sand iris]

-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

[sand iris]

colab is still a thing

https://cdn.discordapp.com/emojis/1104640088840876093.webp?size=48&name=clueless&quality=lossless

[wild mason]

Hello Gdr-!
There is any update for mac? (with apple sillicon)

[sand iris]

for non-realtime ?

[wild mason]

yes

[sand iris]

I don't think training is supported, but for inference :

• Applio should work (doubt on that)
• Mangio RVC should work (but it hasn't been updated in a while)
• Mainline RVC should also work

Not sure on how installation goes, might or might not be manual

[wild mason]

for realtime?

[rare goblet]

@wild mason has reached level 1.

[wild mason]

thank you so much

i just deleted applio rn, it did not work

[oblique nebula]

can rvc disconnected auto-cut the dataset audio?

[proper shale]

in every 4 seconds yes

tho if u wanna trim silences

its better to use audacity

[oblique nebula]

so i have to upload the zip cotaining the full audio?

[rare goblet]

@oblique nebula has reached level 1.

[proper shale]

yeah

if its 1hr of audio separate every 30 mins

[oblique nebula]

why rvc disconnected is training so fast?

[oblique nebula]

agora que vi q vc é br

E A GENTE FALANDO EM INGLES

pq o rvc disconnected ta treinando tao rapido?

[proper shale]

kkkkkk n sabia q tu era br

[proper shale]

cm assim mn

tipo, 1 segundo por epoch?

[oblique nebula]

sim

ou menos

[proper shale]

Pode checar a pasta weights tbm? pode ser q o treino deu errado

[oblique nebula]

deu certo

mas ta muito rapido

[oblique nebula]

me add ai

[proper shale]

manda aq msm

[oblique nebula]

deixa baixo

pergunto pro meu primo

m0dz

[sharp ore]

there is someone fumbling for kaggle?

I have a problem and I do not know what to do about it

[mellow token]

-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

[snow hazel]

So I'm confused on how to make an AI cover if colab/Gradio isn't working.

[rare goblet]

@snow hazel has reached level 1.

[wispy burrow]

Apparently those in #1159949278270193734 are said to be working...

[snow hazel]

Weird cause it isn't working

[modest topaz]

anyone know what happened here? im trying to install all the packages with pip install -r requirements.txt

[wispy burrow]

You have Python 3.11?
Try a different version.

[modest topaz]

what do i need to update it?

[wispy burrow]

Try downloading Python 3.9.8.
That one is recommended for Applio-RVC.
And works for Mangio too.
Which RVC are you trying to use?

[modest topaz]

i have no clue which one is better so i just picked this

[wispy burrow]

It's a bit older, but it's good.
I use it too.

[modest topaz]

alr

im just tryna train a model of my friends voice cuz why not

also this AI stuff seems pretty cool

[modest topaz]

just installed 3.9.8 and this is still happening

[wispy burrow]

I have to say I didn't have to pip install anything for Mangio.
I just had to extract the zip for it to work.

[modest topaz]

well idk im just following the github

[wispy burrow]

Did you enable the "Set as PATH" option?

[modest topaz]

idk

il redo it just in case

[wispy burrow]

I just followed this guide https://docs.google.com/document/d/1KKKE7hoyGXMw-Lg0JWx16R8xz3OfxADjwEYJTqzDO1k/edit#heading=h.e9y3sfgsycb8

And downloaded the zip from here https://huggingface.co/MangioRVC/Mangio-RVC-Huggingface/tree/main

[modest topaz]

dam

[rare goblet]

@modest topaz has reached level 2.

[modest topaz]

it seems to be working now so far

GODDAMMIT

very next command

[modest topaz]

i give up im just gonna do this#

ok wtf that was easy

why did the github try make me do all that

i just had to run 1 single file i already had

[snow hazel]

What's the current version of colab/Gradio?

[wispy burrow]

Delete the previous RVC folder then.
Also, I see the folder named 'RVC Models'.
Remove the space there and name it 'RVC-Models' or something.
It's not recommended to have spaces or special characters in the path to the RVC folder.
It can cause issues otherwise.

[modest topaz]

dam ok ty

how do i make a dataset?

i have a couple clips of my friend just talking

but idk what to do with that

[wispy burrow]

I think this guide will explain it well enough.

RVC v2 Training Guide
https://docs.google.com/document/d/13ebnzmeEBc6uzYCMt-QVFQk-whVrK4zw8k7_Lw3Bv_A/edit#heading=h.bjzhhhcn3f69

[rare goblet]

@wispy burrow has reached level 6.

[modest topaz]

ok so i have this folder with these audio clips of my friend talking for 2-3 mins in each one

is this what it means by dataset?

[wispy burrow]

In short, you don't want background, echoes or reverb.
You don't want silence.
Want around 10-50 minutes of audio.

[wispy burrow]

Samples of one's voice. preferably in WAV or FLAC.
So yes.

[modest topaz]

is this good enough quality

[west cipher]

trying to start applio fork, but theres a ton of missing modules when i start it using the intel gpu/cpu option

had to pip install sklearnex as well as torch modules manually and now its telling me to install something else, is there a way to fix this?

[serene musk]

I'd say 12 minutes of clean speech is just fine / great for a dataset

[modest topaz]

so is the file i sent above a decent example of that?

(ignore the scottish waffling)

[wispy burrow]

have tried pip installing the requirements.txt file(s)?

[serene musk]

If it's clean and has no reverb and cut out the silence as best u can then yeah

[west cipher]

whats the command for it?

[serene musk]

https://youtu.be/ZbAT8IbsitU

[wispy burrow]

The file is in assets/requirements.
So locate them and open the cmd there.
Then pip install -r requirements.txt should work.
@west cipher

[west cipher]

its saying it cant install because only python 3.7>3.11 is supported yet i'm running 3.11.4

unless they mean that it literally can only go up to version 3.11 and nothing after

[rare goblet]

@west cipher has reached level 2.

[wispy burrow]

3.11 is not fully supported.
3.9.8. was recommended for Applio.

[west cipher]

oh ok

well im gonna downgrade to 3.9.8 and see if it works

[sand iris]

3.9 or 3.10 would work

Not sure why y'all insist on using 3.9**.8** though, it's way out of date

[wispy burrow]

It was recommended and linked in the Applio installer.

[sand iris]

I have reasons to believe you can't trust applio

3.9.8 - Nov. 5, 2021

missing 2 whole years of security fixes, woooo !

[proper shale]

yay!!

[ionic pilot]

help

[brittle wing]

['trainset_preprocess_pipeline_print.py', '/notebooks/Mangio-RVC-Fork/datasets/literallythefuckingubisoftintro', '48000', '4', '/notebooks/Mangio-RVC-Fork/logs/thefuckingubisoftsound', 'False']
/notebooks/Mangio-RVC-Fork/datasets/literallythefuckingubisoftintro/.ipynb_checkpoints->Traceback (most recent call last):
File "/notebooks/Mangio-RVC-Fork/my_utils.py", line 134, in load_audio
ffmpeg.input(file, threads=0)
File "/usr/local/lib/python3.9/dist-packages/ffmpeg/_run.py", line 325, in run
raise Error('ffmpeg', out, err)
ffmpeg._run.Error: ffmpeg error (see stderr output for detail)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/notebooks/Mangio-RVC-Fork/trainset_preprocess_pipeline_print.py", line 84, in pipeline
audio = load_audio(path, self.sr, DoFormant, Quefrency, Timbre)
File "/notebooks/Mangio-RVC-Fork/my_utils.py", line 141, in load_audio
raise RuntimeError(f"Failed to load audio: {e}")
RuntimeError: Failed to load audio: ffmpeg error (see stderr output for detail)

how do i fix this

is the audio too short? (5 seconds)

making a model, not inferencing

[fierce rock]

Hey this is probs a super stupid question but where do I put models so I can access them from in inferencing voice drop down?

[proper shale]

Have you unzipped the file?

[wispy burrow]

You didn't install/extract Mangio on your C Drive?

[proper shale]

On the weights folder

Index should be placed in logs, inside a folder with your model's name

[fierce rock]

Thanks that was my next question.

[proper shale]

I read your mind haha

[autumn shard]

Why does one of the weights model's download button leads me to some guy's music video?

[brittle wing]

that error doesn't do anything

it still works if i ignore it

it's just i got another issue now

wait i may have fixed it

[proper shale]

💀 what

what model is this

[brittle wing]

nope nvm

训练结束, 您可查看控制台训练日志或实验文件夹下的train.log

what

[autumn shard]

the Arby's model

[rare goblet]

@autumn shard has reached level 3.

[autumn shard]

you click download and it leads to a music video

[brittle wing]

INFO:torch.distributed.distributed_c10d:Added key: store_based_barrier_key:1 to store for rank: 0
INFO:torch.distributed.distributed_c10d:Rank 0: Completed store-based barrier for key:store_based_barrier_key:1 with 1 nodes.
gin_channels: 256 self.spk_embed_dim: 109
INFO:thefuckingubisoftsound:loaded pretrained pretrained_v2/f0G32k.pth
Process Process-1:
Traceback (most recent call last):
  File "/notebooks/Mangio-RVC-Fork/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 "/notebooks/Mangio-RVC-Fork/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 "/usr/lib/python3.9/multiprocessing/process.py", line 315, in _bootstrap
    self.run()
  File "/usr/lib/python3.9/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/notebooks/Mangio-RVC-Fork/train_nsf_sim_cache_sid_load_pretrain.py", line 200, in run
    net_g.module.load_state_dict(
  File "/usr/local/lib/python3.9/dist-packages/torch/nn/modules/module.py", line 2041, in load_state_dict
    raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for SynthesizerTrnMs768NSFsid:
    size mismatch for dec.noise_convs.0.weight: copying a param with shape torch.Size([256, 1, 64]) from checkpoint, the shape in current model is torch.Size([256, 1, 80]).
    size mismatch for dec.ups.0.weight_v: copying a param with shape torch.Size([512, 256, 20]) from checkpoint, the shape in current model is torch.Size([512, 256, 24]).
    size mismatch for dec.ups.1.weight_v: copying a param with shape torch.Size([256, 128, 16]) from checkpoint, the shape in current model is torch.Size([256, 128, 20]).
    size mismatch for enc_q.pre.weight: copying a param with shape torch.Size([192, 513, 1]) from checkpoint, the shape in current model is torch.Size([192, 1025, 1]).```

[proper shale]

uhm

did you

process the dataset

[brittle wing]

i did

oh wait could it be that error?

the audio is an mp3

5 seconds long

not sure abt sampling rate

[clear hound]

Is there anyway for RVC to have no delay or little to none? I really wouldn’t want to speak and be heard like 20-30 seconds later 🙏

[proper shale]

u gotta process dataset correctly

even if its a 5 sec audio

[proper shale]

First off - what's your GPU? And what settings do you currently have?

[brittle wing]

it doesn't work tho

[proper shale]

How does the dataset folder look? Can you show me?

[brittle wing]

/notebooks/Mangio-RVC-Fork/datasets/ubintro/.ipynb_checkpoints->Traceback (most recent call last):
  File "/notebooks/Mangio-RVC-Fork/my_utils.py", line 134, in load_audio
    ffmpeg.input(file, threads=0)
  File "/usr/local/lib/python3.9/dist-packages/ffmpeg/_run.py", line 325, in run
    raise Error('ffmpeg', out, err)
ffmpeg._run.Error: ffmpeg error (see stderr output for detail)

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/notebooks/Mangio-RVC-Fork/trainset_preprocess_pipeline_print.py", line 84, in pipeline
    audio = load_audio(path, self.sr, DoFormant, Quefrency, Timbre)
  File "/notebooks/Mangio-RVC-Fork/my_utils.py", line 141, in load_audio
    raise RuntimeError(f"Failed to load audio: {e}")
RuntimeError: Failed to load audio: ffmpeg error (see stderr output for detail)```
spits out this error at me

currently doing ubintro

i remade the sound to be 48k sampling rate

[wispy burrow]

Is that folder zipped?

[brittle wing]

no

i changed the mp3 file for a wav, still no change

[proper shale]

dataset path should be

/notebooks/Mangio-RVC-Fork/datasets/ubintro

i think

no slash on the end

try that

[brittle wing]

/notebooks/Mangio-RVC-Fork/datasets/ubintro

that's what i have

same error

[proper shale]

try with the slash at the end

[ionic pilot]

-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

[shy zealot]

I'm sorry, I couldn't find any model with that query you gave me, try with another one.

[brittle wing]

-rvc

[azure marsh]

RVC Guides (how to make ai cover)

🍏 Applio-RVC-Fork: Guide (EN/ES)

https://docs.google.com/document/d/12YmIQFLFzth1YsZJMGrA0aeTN0Go8DNZVw8LFSf-Hug

Tip: You can also watch this video tutorial made by iroaK if you need further help.

RVC v2 Tutorial for Paperspace (By LollenApe)

https://docs.google.com/document/d/1lIAK4Y0ylash_1M2UTTL_tfA3_mEzP0D2kjX2A3rfSY/edit?usp=sharing

RVC Guides in Other Languages

• Russian FAQ for RVC/Voice changer
• Italian Guides or -it_g
• RVC Guides in PT-BR (-rvc_br)

[shy zealot]

Taylor Swift | 1989 Era

Upload the <t:1691292195>
Server uploaded: AI HUB
Channel uploaded: AI HUB Post
Profile Owner: winwln

Author

wlnwin

Discord User

winwln

Internal ID

3444

Language

EN 🇺🇸

Type

RVC V2

Size

171 MB

Extractor

RVMPE

Epoch

400

Steps

0

Powered by ModelAI.net, Made with Love ❤️

Melanie Martinez Portals era

Upload the <t:1688522357>
Server uploaded: AI HUB
Channel uploaded: AI HUB Post
Profile Owner: jaguleirah

Author

jaguleirah

Discord User

jaguleirah

Internal ID

11930

Language

EN 🇺🇸

Type

N/A

Size

0 MB

Extractor

N/A

Epoch

500

Steps

0

Powered by ModelAI.net, Made with Love ❤️

Selena Gómez

Upload the <t:1685905099>
Server uploaded: AI Hispano
Channel uploaded: AI Hispano Post
Profile Owner: aiverse

Author

AIVERSE

Discord User

aiverse

Internal ID

187

Language

EN 🇺🇸

Type

RVC

Size

0 MB

Extractor

null

Epoch

1000

Steps

0

Powered by ModelAI.net, Made with Love ❤️

[light orchid]

this message keeps happening when i post a model to try to get the role, "Invalid file size for "image". Please attach a file that's under 25 MB." My image isn't above 25MB, and it's a right type, what does it mean, invalid size?

can someone help me or validate my model?

[quiet trail]

how do you download rvc

[daring verge]

theres a guide for that

-rvc

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English Documentation

https://us.aihubfrance.fr/

Credits: AI HUB France

🇺🇸 English FAQ

https://us.aihubfrance.fr/faq

Credits: AI HUB France

Guides

🍏 Applio-RVC-Fork: Guide

Link: Google Docs

Credits: IA Hispano

How to install RVC on windows

Link: YouTube

Credits: iroaK

RVC v2 Tutorial for Paperspace

Link: Google Docs

Credits: LollenApe

🌐 Translations

• 🇫🇷 French Documentation
• 🇮🇹 Italian Guides or -it_g
• 🇷🇺 Russian FAQ for RVC/Voice changer
• 🇧🇷 RVC Guides in PT-BR (-rvc_br)

[bold ermine]

@brittle wing does your error resolved?

[quiet trail]

i open it and put the voice i wanted and it aint working

[rugged stirrup]

anyone know how to install tensorboard

[modern snow]

Does anybody have suggestions for making ai models?

[brittle wing]

can i get a tutorial on how to use this new colab for rvc2? so confusing

[rugged stirrup]

does anyone know what: The system cannot find the path specified.
The system cannot find the path specified.
The system cannot find the path specified.

PACKAGES INSTALLED
Press any key to continue . . .

means cause I am trying to follow the guide to install ROD

[brittle wing]

can anyone tell me how to quit the local on mac M1 ?

i wanna restart RVC but can't seem to kill the localhost

[proper shale]

I think you should close the Terminal

That's how it works on Windows at least

Should be the same, I believe

[brittle wing]

it doesn't work.. i did this many times...

[proper shale]

oh damn

[brittle wing]

its like since yesterday.. i can't get it to stopp

[rare goblet]

@brittle wing has reached level 1.

[brittle wing]

-it_g

[azure marsh]

Risorse

• Link utili a cura di Ilaria

https://rentry.org/link-utili

[proper shale]

Check Activity Monitor

And see what processes are related to RVC - i think you will be able to kill it that way

[rugged stirrup]

could anyone help with installing tensorboard I get this error when trying to run requirements.bat: The system cannot find the path specified.
The system cannot find the path specified.
The system cannot find the path specified.

PACKAGES INSTALLED
Press any key to continue . . .

[noble latch]

Which virtual cable is the best?

[ionic pilot]

help

[glad zealot]

did you run the feature extraction cell?

[zinc anchor]

What does your dataset look like? Were you able to load it without issues?

[ionic pilot]

yes

[glad zealot]

it says no feature in the error so not sure

[glad zealot]

can you show the console of the feature extraction cell?

[zinc anchor]

You should upload your dataset as a zip file and load it that way

[ionic pilot]

Can I show you how to do it live? Do you have a moment?

[zinc anchor]

You just need to compress that folder into a zip and reupload it

[main patio]

Is there a tutorial for RVC disconnected?

[glad zealot]

show the console of this part

it might have given an error

[zinc anchor]

Not a video tutorial but for now there's this guide by Ange https://docs.google.com/document/d/1XuxQYiqEhYrdYeCZRRLrmV_ciMKo0bV-jTCGHu_-5Cc/edit?usp=drivesdk

[rare goblet]

@ionic pilot has reached level 2.

[zinc anchor]

We're working on a more detailed guide though, idk when it will be done

[glad zealot]

video seems to be easy to do just annoying to record lol

[zinc anchor]

Yeah ig

[ionic pilot]

I can't compress to zip

[glad zealot]

compress it before addint it to your drive

[zinc anchor]

Then you need to install a program like 7zip

I assume you have the dataset folder on your PC

For 7zip you can right-click on the folder and choose "add to foldername.zip" in the 7zip actions there

[zinc anchor]

Yeah because when you load the dataset you need to give a zip file's name to the "load dataset" cell so that it can look for it in the RVCDisconnected folder and extract the dataset from there in the way that the notebook wants

[glad zealot]

you can technically zip it

[ionic pilot]

okey i compress

[glad zealot]

in colab

[ionic pilot]

done

[fierce hollow]

hey i really need some help . i was training a voice model with 750 epocs but at 670 something the gpu got disconnected in the google collab and i wanna know can i resume it with the log files saved in drive ?

[ionic pilot]

All right?

[fierce hollow]

or am i supposed to start over again ?

[zinc anchor]

Yeah, you can upload that to the rvcDisconnected folder and re-run all the cells from the pre-processing section (give zabson.zip for the load dataset cell)

And then you can train hopefully

[fierce hollow]

okay

[zinc anchor]

Sorry I'm talking to @ionic pilot

[fierce hollow]

oh mb

[ionic pilot]

thanks bro

❤️

[ionic pilot]

and u thanks too

[zinc anchor]

I think that potentially you can resume training

[fierce hollow]

i have to resume in a different google account

[rare goblet]

@fierce hollow has reached level 1.

[zinc anchor]

I haven't tried myself yet

[fierce hollow]

okay i will try

[zinc anchor]

I think you'd have to download the rvcDisconnected folder with everything in it and upload it on your other account's Drive

[fierce hollow]

okay

okay i think it worked

so how many epochs should i give now ?

it was like 670 something when it ended so i have if i give 30 in training will the output be 700 epocs ?

[zinc anchor]

I'd input 700

[zinc anchor]

Have you imported the model though?

[fierce hollow]

okay but will it take all the time or will it take time of 30 epochs

[zinc anchor]

Check the number of steps

[fierce hollow]

yeah

[zinc anchor]

It's probably 2333333 due to the "save_only_last_checkpoint" thing

[fierce hollow]

i just left it defualt

[zinc anchor]

In the old notebooks it understood that you were at 670 epochs and continued the training from there and finished at 700

[fierce hollow]

should i change it to 2333333 ?

[zinc anchor]

If it does more than that you can stop it

[zinc anchor]

Look at your G and D files in the model folder on Drive

[ember crow]

put the number after G_ and D_

[zinc anchor]

Ye

[fierce hollow]

where can i find it ?

[zinc anchor]

In the rvcDisconnected folder

[fierce hollow]

i can find a zip folder logs

[rare goblet]

@fierce hollow has reached level 2.

[ionic pilot]

@zinc anchor experiment name zabson or zabson.zip

[zinc anchor]

zabson there, zabson.zip in the "load dataset" cell

[ionic pilot]

okey

[zinc anchor]

Wait you're supposed to have a model folder there

After the training you did

I think you did not save the preprocessed files the first time, or did you do that?

[ionic pilot]

I'm doing everything from scratch

okey its good

the fault was that it did not search for the folder as it is in the .zip it searches

[zinc anchor]

Oh then in this case you are starting over from 0 epochs

[rare goblet]

@zinc anchor has reached level 11.

[ionic pilot]

yes but thanks for the help because I've been struggling with this since yesterday

❤️

[zinc anchor]

Wait sorry, I'm mixing you and the other guy up

Yeah you should be fine now @ionic pilot

[ionic pilot]

yes thanks

[fierce hollow]

if u were talking to me i think i have a model folder

in that i have rvclogs.zip folder

[main patio]

Is there a way to download AI covers online without downloading any software?

[zinc anchor]

No G and D files? I'm starting to think that the "resume training" option is for when you can finish training and save the files but then you want to keep training that model after it's finished

So you probably have to start over unfortunately

[fierce hollow]

okay sad

[zinc anchor]

I'd set lower epochs this time to avoid getting disconnected before the training is over

Depending on how long your dataset is

[fierce hollow]

how many epochs should i go for 20 min dataset

the vocals are like really good

[ionic pilot]

tensor dont work

[zinc anchor]

Between 300 and 400 should be fine

[fierce hollow]

okay i will go with 350 then

[zinc anchor]

Oke

[fierce hollow]

anyway thanks man

[zinc anchor]

Yeah it happened to me as well the only time I trained a model on that notebook

We have to tell the person who made it

[zinc anchor]

You can train anyways, you just won't be able to see all the metrics

[unreal pendant]

google colab is back? no more bans?

[glad zealot]

you can use rvcDisconnected

[stone dawn]

wish there is working rvc collab working interface ai covers

[glad zealot]

lucky for you there is :D

for covers that is

[stone dawn]

ommg whereeeee

[glad zealot]

are you gonna do covers?

[stone dawn]

yeahh

[glad zealot]

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

Do note that WEB UIs are technically still banned on colab

The colab might not work in the future due to Google finally taking it down, and there is a chance that your account might also get banned in the process

So use at your own risk (or use a throwable alt)

[stone dawn]

omgg thankk youuuu 🙏

[unreal pendant]

i am using applio anyway, but i can inform users on my server if any colab is working with training

[glad zealot]

yup

[unreal pendant]

this one works, right?

[glad zealot]

thats for cover making and yes

[stone dawn]

is it the same way to use?

[glad zealot]

yup

same as the old one

[stone dawn]

ah thank youu!!

🫂

[glad zealot]

https://cdn.discordapp.com/emojis/1091211052458455111.gif?size=48&name=HeadPat&quality=lossless

[zinc anchor]

Colab has always been there, they just made it impossible to use the notebooks with a WebUI

For now we have a couple notebooks working but they don't have a WebUI (which is what Gradio was), it's all done through the cells

[unreal pendant]

understandable

[rare goblet]

@unreal pendant has reached level 1.

[unreal pendant]

thanks

[glad zealot]

not quite impossible.. just very annoying

[zinc anchor]

Yeah

I can still use the original AICoverGen as a Pro user but that's it

[glad zealot]

use it without pro :D

[zinc anchor]

Doesn't it disconnect you after 10 minutes?

[glad zealot]

nope

[zinc anchor]

Oh nice

[glad zealot]

lol been 2 hours

[zinc anchor]

Bruh

[glad zealot]

its my mod

[glad zealot]

dis

[zinc anchor]

Ooooh

[ionic pilot]

work on operagx

google dont work

[zinc anchor]

Interesting

[unreal pendant]

[tight leaf]

who has an open repository of this project? - RVC_V2_Huggingface_Version

[tight leaf]

HOW?! ITS WORK

[glad zealot]

yes it works

[tight lava]

I keep getting this error

[rare goblet]

@tight lava has reached level 1.

[tight lava]

"'NoneType' object has no attribute 'setdefault'"

[glad zealot]

show me

on what cell?

[tight lava]

When I press generate

It downloads the song then says that error

https://colab.research.google.com/github/SociallyIneptWeeb/AICoverGen/blob/main/AICoverGen_colab.ipynb

This is the one I use

[glad zealot]

ye that one doesnt work

[tight lava]

Ohh

[glad zealot]

if you're using the web ui version

[glad zealot]

use this for web ui

[tight lava]

Kk thanks

[tight leaf]

God, you're brilliant at using ROT13 to encrypt code, but can you do it with more convenient repositories for creating AI covers?

[rare goblet]

@tight leaf has reached level 1.

[glad zealot]

it doesnt work most of the time

just got lucky that it works with covergen and voice changer

[pulsar snow]

Wow! So that doesn’t disconnect?

[glad zealot]

nope

well after around 4 hours or something it will

which is normal

[tight leaf]

damn, that is, you can't do this with another repository? it's just not particularly convenient

[pulsar snow]

Yeah, that’s alright. It was happening before! Thank you so much! I’ve been struggling with making ai covers because it kept disconnecting

I had to speed run 😂

[tight leaf]

I just liked the old repository from mangio for creating ai covers

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing#scrollTo=7vh6vphDwO0b

[pulsar snow]

Thanks! I’ll try both later today!

[rare goblet]

@pulsar snow has reached level 1.

[pulsar snow]

I thought we had lost to Google

[silk ferry]

Wow

[rare goblet]

@silk ferry has reached level 3.

[clever crest]

https://app.kits.ai/
Question if i want to make a female cover song using a female ai voice do i keep the octave change to 0?

Since from what I learnt octave change needs to be highter if you convert female ai voice for male covers

[karmic oracle]

Im getting and error that says
Traceback (most recent call last):
File "D:\RVC-GUI-pkg\rvcgui.py", line 129, in vc_single
audio_opt = vc.pipeline(
File "D:\RVC-GUI-pkg\vc_infer_pipeline.py", line 340, in pipeline
self.vc(
File "D:\RVC-GUI-pkg\vc_infer_pipeline.py", line 243, in vc
(net_g.infer(feats, p_len, pitch, pitchf, sid)[0][0, 0])
RuntimeError: CUDA error: the launch timed out and was terminated
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Any help

I tried to do the oof and on solution it didn't work

[rare goblet]

@karmic oracle has reached level 1.

[karmic oracle]

Off**

[proper shale]

Yeah keep it zero

[clever crest]

thanks

[rare goblet]

@clever crest has reached level 1.

[junior zenith]

Hi!

[modest topaz]

this happends when trying to train my model?

[rare goblet]

@modest topaz has reached level 3.

[glad zealot]

i guess? so you can train on one and play on the other

[proper shale]

4GB VRAM won't cut it for training

[proper shale]

...well actually

Batch_size is applied evenly with both GPUs

[modest topaz]

bruh

[glad zealot]

ohh cool

[modest topaz]

so i physically cant do it then?

[proper shale]

If you try like 2 batch_size then maybe

[glad zealot]

btw WHY ARE PEOPLE DM-ING ME

AAAAAAAAA

[proper shale]

But I'm not sure

[proper shale]

Happens all the time sadly

[proper shale]

Welp

I guess you just can't do it physically then

[modest topaz]

bruh

[nocturne ferry]

how much epoch would ya recommend for 37 minutes of voice footage

[proper shale]

Yeah sorry

[proper shale]

Varies a lot

The graphs will tell you when you don't need more epochs

[nocturne ferry]

i see thank you

[ornate solstice]

Hi, so i have an amd gpu (very powerful one) but it the sound is being translated by the ai very slowly at speeds like 11s or 20s

what can i do to minimize it?

[proper shale]

What are you using exactly - the voice changer?

[tight leaf]

help , how to fix it?

Error
Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx

[proper shale]

...You need an NVIDIA driver

Do you even have a GPU

[tight leaf]

no I'm doing notepad via google colab

[proper shale]

Oh, you need to change your runtime type

Click on the top right thing with the stats

[tight leaf]

ааа..

[tight leaf]

and could you also help me with google collab, the fact is that the code cell takes up a lot of space when you run it, I want to hide the text somehow

[proper shale]

thats the output

you cant really hide it

oh wait nk

if u double click it might go away

[west cipher]

..so basically I have an acapella for Where Are Ü by Skrillex and Justin Bieber, I want to make an ai cover out of it but because of the large amount of reverb and echos I feel it’s gonna be pretty difficult

[proper shale]

There are options to deverb stuff

check MVSep

Ultimate Vocal Remover --> UVR-DeEcho-DeReverb

[west cipher]

Thanks, do you know what would be a good aggressiveness level for that song

for the dereverb

[proper shale]

try 0.5

if it isnt enough go for like 0.8 (but this will decrease quality though!)

[chilly summit]

whats the best pitch extraction for training? is it rmvpe?

[proper shale]

rvmpe for basically 90% of use cases is better

so yea

[west cipher]

can someone help me with this again

so basically in order to start applio i was on an unsupported version, i downgraded, reinstalled the requirements, now its back to saying this again

[proper shale]

oh those modules

you have to do pip installs for those

search for the best way to install those

cuz pip install sklearnex could work but also might not

[west cipher]

but i did pip install -r requirements.txt and it should have installed them

[proper shale]

thing is, they probably dont install correctly

Applio has way too many libraries and sometimes the order in which they are downloaded screw up other modules

[west cipher]

okay so i figured out the command

basically sklearnex is short for scikit-learn-intelex

[proper shale]

oh

[west cipher]

so when i did pip install scikit-learn-intelex

it worked and its installing

okay im gonna try starting up applio again

[slender lotus]

can someone link me an up and running google colab that doesnt shut off after two seconds?

i dont see a correct link anywhere posted to actually use my model

[proper shale]

#1159949278270193734

[slender lotus]

thank you!

[pliant ingot]

are there any web ui rvc

[west cipher]

after about a half hour of installing, looks like applio has started and is working fine.

thanks mj

[median ivy]

help

[ember ermine]

go to the Set Training Variables section and name your experiment something

[proper shale]

np

[proper shale]

^ @median ivy

make sure you have ran the step too

[median ivy]

ok

[meager quarry]

I keep getting "error: credential propagation was unsuccessful" while trying the step Mount Drive in RVC v2 Disconnected colab

[proper shale]

hm, could be a problem while logging in drive

or u didn't give colab perms

[median ivy]

what

[rare goblet]

@median ivy has reached level 1.

[meager quarry]

I also had a similar error just now

is the colab itself broken somehow?

[median ivy]

idk

[meager quarry]

Just had the same error as waddle too again

[slender lotus]

i feel like they're all broken

[rare goblet]

@slender lotus has reached level 1.

[slender lotus]

none of the links provided in #info are working

[sand iris]

did you restart the runtime ?

[meager quarry]

Tried that myself, still same error as in screenshot

[rare goblet]

@meager quarry has reached level 1.

[jaunty shale]

No matter what type of voice I try to train locally, it has the same type of voice. I put 4 CPU Threads, rmvpe-gpu, saving frequency to 10, epochs to 300 and batch size 4 or 5, it always ends up with steps having the same amount as epochs.

Harvest doesn't do that, but that one is extremely slow. Rmvpe is much faster but results aren't good.

Intel Core iI5 10300H

Nvidia GeForce GTX 1660 Ti
Overall Memory: 13,9 GB
Dedicated Memory: 6 GB

Does anyone know how to fix this issue? I've been trying to train a voice for nearly a month now and no one tried to help :(

[meager quarry]

Did you train that index thing beforehand

[median ivy]

i restarted before

[jaunty shale]

yeah I always train the index before training a model

[meager quarry]

weird, can't help unfortunately

[jaunty shale]

I could try and show that it does that rq

[rare goblet]

@jaunty shale has reached level 1.

[jaunty shale]

actually no, I'll try downloading the latest version of RVC

[ornate gulch]

hi

is the rx580 8gb good for training models ?

[meager quarry]

so basically disconnected colab is probably broken, so online model training doesn't work, fantastic

[west cipher]

I have a zip file to a model that was originally on weights but was deleted a couple months ago, all thats in the zip file is the .pth and .index file, how do I load it into applio?

figured it out, just had to click Refresh for it to load in.

[slender lotus]

can someone link me a currently working google colab? all the ones i tried in ⁠📘┋info did not work 😦
or a huggingface link

[meager quarry]

same issue here

[pulsar snow]

Check the pinned one

[meager quarry]

too bad that one is for covers only appearently

is there a working training colab somewhere

[mossy pasture]

the discontinued v2 works? i just trained a model out of it, but you need to understand how to use it, which is a little complicated

[compact wraith]

I always thought rmvpe would be the best quality,

and i get that mangio is also good as well, it doesnt really come off as good, sometimes crepe almost works for some days

[jaunty shale]

Well I used Retrieval-based-Voice-Conversion-WebUI and instead of crepe I had rmvpe and rmvpe-gpu.

For some reason it went fast, but steps were same value as epoch (e300_s300) for instance

[ornate gulch]

should i buys the gtx1060 3gb or rx580 8gb , which is better ? please

[meager quarry]

I know how to but I just keep getting
when running the first step: "fatal: destination path 'Mangio-RVC-Tweaks' already exists and is not an empty directory"
but I have nothing of it installed on my google Drive

[mossy pasture]

it's more of the runtime problem instead of your google drive problem

just restart a new runtime

[compact wraith]

i remember there was rmvpe_onnx

i think, im not sure what that does, but i didnt really get into that program

[rare goblet]

@compact wraith has reached level 8.

[compact wraith]

it does run pretty good

[jaunty shale]

that's for realtime, I talk about local voice training

[meager quarry]

same error. I restarted the runtime and the same thing happened again

[rare goblet]

@meager quarry has reached level 2.

[mossy pasture]

maybe us another account?

it's rare to see error when installing dependencies

[compact wraith]

i use colab for training, way different than using local,

because i tried it local, and somehow it dont work

[jaunty shale]

isn't colab one banned?

cuz of google

[mossy pasture]

nah, the discontinued v2 still works, but you just don't have gradio

and for inferencing, since it doesn't take much gpu usage, you can just do it local

[compact wraith]

RVC Crepe v2 colab still works, but you basically have to pay units to run

i think it doesnt work at all for free users

[meager quarry]

But why did it happen in the first place? I'm clearly not banned from colab nor is my google account compromised.

[brittle wing]

does it not anymore?

[rare goblet]

@brittle wing has reached level 1.

[brittle wing]

it used to

[meager quarry]

it keeps saying "error: credential propagation was unsuccessful" when I try Mount Drive step too

[meager quarry]

And this is what happens when I try to click on the prompt to give permission to Google Drive to make changes in the first place

[compact wraith]

for free users wont work, but for paid users, well for most of the time yes

[brittle wing]

idk it used to

unless they changed it

[meager quarry]

that's not the disconnected rvc v2 one right

[compact wraith]

nah the RVC Crepe v2 + persistent storage / autosave one, kind of the old one

[meager quarry]

oh that one

[compact wraith]

i think so, i just use colab for train, local for conversions, or track isolations

[meager quarry]

Anyway, when I am not even able to give permission to google drive in the first place and get an error only when I try, why shouldn't I assume the disconnected v2 colab itself is broken somehow

[compact wraith]

i probably need a tutorial for that because i dont know how that thing works

[west cipher]

this isnt really rvc related but do you guys know how to download kits.ai models, i made my own kits model a while back but now i want to download it and use it in the collab

[meager quarry]

Nah, this was more so directed at people who claim the colab works rn, not at you specifically
I'm not even the only one who get these errors today

[compact wraith]

I only have one trash hardware system, so, it wont really work as i was expecting it to be

[meager quarry]

I'm just more confused as to why so few people are noticing this, I mean out of 18k members there have to be 50 people at least trying to actively make models

[rugged stirrup]

does anyone know how to install just normal tensorboard cause real time overtraining detection doesnt work for me

[meager quarry]

And why is there no announcement or pin for this anyway? This sounds like newsworthy to me.

[wispy burrow]

Installing TensorBoard Locally for RVC (without errors)
https://docs.google.com/document/d/1ihvqWnSyabvZ9jX6BsOxEY1FpplwjY3Jse3she_VO6I/edit

[compact wraith]

well, when this server started running, i guess that Model AI bot wasnt running,

so people thought they lost all the models that was from the og server, but its been archived now

they did made a annocement for that

it would be hard for all staff to do it manually, since it will take forever

[rugged stirrup]

I keep getting environment errors

[wispy burrow]

Some models are lost, but the majority can still be found I think.

[meager quarry]

So I'm guessing by the radio silence here that rvc disconnected is actually broken right now. I wish there was an alternative to it that allowed one to download models.

[compact wraith]

Probably. i think they took out all of the RVC v1 models, like the very outdated ones, maybe some v2 models and the recent ones,

probably 1 month or 2 month old is still availaible

[meager quarry]

The only other ""alternative"" being cpu training locally to only get 25 epochs in like 5 hours and waiting around like a total idiot

[wispy burrow]

25 minutes for 1 Epoch on CPU. Yeah...

[meager quarry]

That's at least what happened when I tried it with rmvpe and a 10 minute dataset, you know, basic requirements for a decent model

[rare goblet]

@meager quarry has reached level 3.

[wispy burrow]

Isn't it requiremenrs.txt?
instead of requirements.bat?

[smoky bough]

Tried a couple versions of RVC V2 and they both keep giving me errors when I try and use it, but it gives different errors each time, most recently being this one

Converting...
Traceback (most recent call last):
  File "/home/user/app/infer.py", line 146, in vc_single
    if_f0 = cpt.get("f0", 1)
AttributeError: 'NoneType' object has no attribute 'get'

How would I go about fixing this?

[rugged stirrup]

if I open install.txt it just opens it in notepad

[brittle wing]

not overtraining right?

[graceful kite]

hi just started using this may i ask why i can't hear anything from output? im using a file as an input

[meager quarry]

don't think so

[rugged stirrup]

how did you get that

[brittle wing]

tensorboard?

[smoky bough]

I'm stupid- I'm using RVC V1 by accident

[rugged stirrup]

yes

[wispy burrow]

Have you tried running that bat file?
What's it for? Mangio doesn't have that...

[brittle wing]

im using paperspace (paid)

[sand iris]

Yeah I'm not sure what that bat script is

[rugged stirrup]

im using it for mangio

[rugged stirrup]

oh

[sand iris]

Which documentation recommends doing things this way ?

[wispy burrow]

Is that for Realtime Overtraining Detection?

[rugged stirrup]

yeah cant find a normal version of tensorboard

[wispy burrow]

I gave you a link to a guide on how to install Tensorboard (not R.O.D.) locally.

[rugged stirrup]

i put the real time overtraining detection in a file inside mangio didnt work and created one and copied the files over to it still didnt work

[wispy burrow]

I think it's this.
https://github.com/grvyscale/AutomatedOvertrainingDetection

[sand iris]

https://cdn.discordapp.com/emojis/863104661342781481.webp?size=48&name=lol~4&quality=lossless

[rare goblet]

@sand iris has reached level 13.

[sand iris]

I don't know what to say honestly

Just funny how messy the RVC installation shit we've created is

[meager quarry]

Alright, I just tried that but got the same error

[torpid cave]

have we gotten an alternative that we can make models with besides kits.ai?

[meager quarry]

So I think I can confidently say the disconnected colab is broken somehow

[meager quarry]

No idea, but I do love how everyone who asks questions like this is just completely ignored.

[sand iris]

what happened to it ?

-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

[meager quarry]

"Mount Drive" doesn't work

when trying to click the prompt to give google drive permission to make changes the error message "error: credential propagation was unsuccessful" pops up

[sand iris]

works here

[meager quarry]

Do you happen to use colab pro?

[sand iris]

No, this is done on my main google account

No subscription whatsoever, this is using the free t4 runtime

[meager quarry]

Then I have no idea why the exact same thing happened on both accounts I tried this on.

[sand iris]

People have been having a lot of issues with it today

not sure why

[rugged stirrup]

does anyone know a instant dataset maker type thing

[meager quarry]

When I restarted the runtime the error changed to on the first step: "fatal: destination path 'Mangio-RVC-Tweaks' already exists and is not an empty directory"
but I have nothing of it installed on my google Drive

[west cipher]

can someone please tell me how to download voice models from kits.ai?

i asked earlier but my message got deleted

bc it was edited

btw the reason why i ask is because i made some good ai models on there before the whole change went through with people quitting kits.ai, i don't want to have to make it all over again if theres a way to just download them

[rare goblet]

@west cipher has reached level 4.

[nocturne ferry]

should rvc be taking hours to finish creating a voice?

[west cipher]

no it shouldnt

but overall it depends on your ram and internet connection

[nocturne ferry]

i see i guess my wifi is bugging or smth i dont have the connection thingy on my pc but i can still use the internet like normal

[meager quarry]

Ok, I want to know what the hell is going on. Why does Colab work for certain people but not for me no matter what account I try? There MUST be something wrong.

Talking about the disconnected V2 colab.

[west cipher]

not sure but i have been using the pinned one since earlier this afternoon

its been about 2 hours and it hasnt stopped

[meager quarry]

But that one is inference only, right?

[west cipher]

yes

but tbh i used to have the same problem as you with it cutting out every couple of minutes

[meager quarry]

I specifically meant the non-webui lemonfoot training colab

[meager quarry]

nono, this isn't like cutting out, it straight up won't mount to drive

[rugged stirrup]

how do you split audio into 10 second clips using audacity using the audio labeler

[west cipher]

Does anyone know what UVR model is used in the Hina Mod ai collab to separate lead/bg/isolated vocals?

the model that the ai uses in the collab to separate vocals

[proper shale]

Voc_FT + MDX B Karaoke

[bronze creek]

how to delete models in rvc

[prime talon]

delete the model logs inside of ./logs/ and delete the model's .pth file(s) inside of ./weight

[rugged stirrup]

whats all-feature-382 mean

[west cipher]

this is the error that i get when i try loading the dataset

[proud elbow]

not yet, but you'd better also see the curve in 0 smoothing, if it spikes down to below 10 and rapidly goes back, can be said it's overfitting (mode collapse).

[glad zealot]

Hmm weird it says found but also not found

[west cipher]

EXACTLY

i honestly dont get it

[glad zealot]

Have you tried reloading the colab?

[west cipher]

Yes

[glad zealot]

That's the latest rvc disconnected right?

[west cipher]

i just clicked on the rvc v2 link in one of the channels here so i think so

[glad zealot]

Ye not sure then

[west cipher]

is there anything else I could use to make a model?

[glad zealot]

Local if you can

[west cipher]

which local

[glad zealot]

The latest rvcdisconnected is in #1159949278270193734 btw

[glad zealot]

The applio iirc

[west cipher]

ohh

so thats why it doesnt work

i was using the old rvc v2

disconnected v2*

wait

nope i was wrong nvm

i have applio but how do i make a model with that

[glad zealot]

I have no idea about that.. I only use disconnected

[rare goblet]

@glad zealot has reached level 14.

[rugged stirrup]

does anyone know why RVC V2 cant do longer than 20 minute datasets or atleast for me it doesnt train the index file but it does the model file fine only V1 works with V1

[west cipher]

RIP me then

[glad zealot]

No idea.. I never made a 20min dataset myself

[rugged stirrup]

oh

[west cipher]

there should be something other than v2 that I could use to make datasets with because this isnt working at all

[rugged stirrup]

has anyone had this issue

[proud elbow]

VRAM limitation perhaps

[rugged stirrup]

it doesnt really use VRAM much

[proud elbow]

my latest Applio works tho

[rugged stirrup]

applio is impossible to set up

[glad zealot]

Lol

[rugged stirrup]

and just errors out and gives me no added index

[ionic pilot]

-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

[proud elbow]

well now I'm trying to train using my old mangio, it works though with 14 mins dataset. I remember I had NaN loss values issue when having an old 6 GB GPU, but now with a 12 GB GPU it works.

[glad zealot]

Sadge my mod not there

[rugged stirrup]

old mangio?

the one with one click training?

[modern snow]

I need help I don't know how to use disconnected

[rugged stirrup]

yeah it works with small datasets for me but I wanna do a 20 minute one

[proud elbow]

I mean (probably) the latest mangio v23.7

[lunar needle]

The tensionboard is not working rn on rvc disconnected

[rugged stirrup]

it is

[lunar needle]

this is colab btw

[glad zealot]

Try reloading the colab tab

Then wait it should show it

Might..

[lunar needle]

Nope, its still there

[glad zealot]

Ig just watch the epochs... Idk if you can see the values tho

The graph basically but in numbers

[bronze creek]

Is there any way to hide cmd and the program from the taskbar?

[quiet haven]

Did you upload the audios correctly?

[west cipher]

i joined them with audio joiner and then it uploaded just the one file as a wav

[quiet haven]

Because this is an error, it is because he was unable to find the audios in the folder

Upload using hugging face instead

[west cipher]

wait do i have to specify a folder?

[quiet haven]

No

[west cipher]

oh ok

[quiet haven]

I will fix this tomorrow

[west cipher]

whats it called on hugging face

i mean like which one on hugging face

[quiet haven]

You have to upload the dataset on hugging and grab the download link to put on colab

[west cipher]

oh okay

[west cipher]

uhh one more thing

i did the huggingface thing and thats what happened

i linked it specifically to the audio

[modern snow]

I'm training a model using v2 disconnected and when it's training it is always saying it is at 0%.

[quiet haven]

Your using the old version, restart the page

[worldly walrus]

is rvc supposed to take up ~10 gigs while you're using it?

[quiet haven]

Is not a error, is normal, the code don't use the normal loading that we are accustomed

[quiet haven]

Depend, if you are training model, yes

[modern snow]

Ok thanks

[worldly walrus]

all i do is generate audio

[quiet haven]

It may be normal depending on the version you have installed

[worldly walrus]

i have the one that first introduced mangio crepe into the ui

but i haven't updated since then

[quiet haven]

So it's normal

[worldly walrus]

ohhh

alright. thank you!

[modern snow]

Alright I'm done with training now what do I do with the files that were dumped into my drive?

[rare goblet]

@modern snow has reached level 3.

[modern snow]

What do I do with the new files in my disconnected folder in my Google drive now I'm done training?

[quiet haven]

Grab the "added" index and the model pth, and zip them both

And your model will be ready

[modern snow]

Ok

Which one of these files as there are multiple pth files

@quiet haven

[west cipher]

rip not again

[upbeat ice]

https://huggingface.co/Hamiltek/ChunLi/tree/main

so can anyone tell me how to get that voice model?-

[quiet haven]

You are using it the wrong way, because the code was running correctly on mine, and this is an error because the audio was not found, are you sure the audio is in wav and the hugging link is resource/main?

[rare goblet]

@quiet haven has reached level 4.

[static gyro]

https://www.amazon.com/ROG-Ally-Handheld-Processor-Microfiber/dp/B0CDL142M9/ref=sr_1_2?crid=2TLOFH07KZGV9&keywords=asus+rog+alloy&qid=1696996589&sprefix=asus+rog+ally%2Caps%2C268&sr=8-2 could this make voice models localy

[quiet haven]

Theoretically yes, as it has a supported video card, processor and you can install Linux, but no make sense to buy a handheld console for training models

[static gyro]

bad structure if it only has the .pth file.

this button

[upbeat ice]

so it's impossible to get it eh?

[static gyro]

wel .index is not needed but it makes it better

[upbeat ice]

welp I've tried it and it gave me this error

[static gyro]

never tried moble

[modern snow]

Alright I have my zip file for my model ready what can I use to test it?