#✨│ai-help

1 messages · Page 41 of 1

rare gobletBOT
#

Ayo? @robust quartz level 1 !!! lfg

proper shale
#

@ = at

robust quartz
#

thank you

brittle wing
#

Hey guys! Just got back to doing AI covers again, is UVR still the best to extract vocals? And what is the most used COLAB for training and doing covers as of now?

finite galleon
jagged elm
#

.ny file is for what?

wide ingot
#

I've installed the ROCm build of PyTorch to test local training, and Mangio-Crepe extraction is so slow.

#

It seems that it may not be suitable for my use case.

flat harness
#

why i sound so laggy in discord?

rare gobletBOT
#

Ayo? @flat harness level 1 !!! lfg

sand iris
wide ingot
#

5500 XT

sand iris
#

Careful, half precision support might or might not work

wide ingot
sand iris
#

supposedly, 2bytes floats (fp16/half precision) compute was broken for a 5700xt user

#

So stuff that deals with this data type might or might not break

#

Unsure about it, I don't have rx 5xxx hardware to test on

jagged elm
#

where to put .ny files guys

sand iris
#

the models they made came out bugged

jagged elm
#

help pls

wide ingot
#

Have to learn the hard way to find out.

sand iris
#

also isn't the 5500xt a 4GB card ?

#

Unlikely that it'll train

sand iris
#

never heard of em

jagged elm
#

npy

#

idk

#

look

wide ingot
rare gobletBOT
#

Ayo? @wide ingot level 11 !!! lfg

wide ingot
#

So training should work.

sand iris
#

Oh, fortunate

jagged elm
#

total_fea.npy

sand iris
#

for some reason someone zipped them in a model

#

the file is useless

wide ingot
#

It's training, although I have to set the batch size to 2 or even 1. This just a few seconds of audio.

wide ingot
#

However, I used RMVPE due to its faster training and inference.

#

I suppose it's something in that user's end that causing problems.

sand iris
wide ingot
sand iris
#

Good that it works though

wide ingot
#

I won't be using it for large datasets though. That'll go through Paperspace.

#

The 5500 XT is simply too slow for that.

wheat rivet
#

I'm using Applio and I finish training my data but I can't seem to get my index file. I press the train feature index and output info shows nothing

molten pecan
wheat rivet
brittle wing
#

does anyone know how to fix this? i already install tensorboard but it still doesnt work

image.png
lean bane
#

what do i do to make the vocals sound like their actually apart of the song? every vocal I've done sounds like another voice pasted onto the instrumentals, but i've heard a few ai covers that sound like proper songs.

lean bane
brittle wing
#

also when i tried to open it it just crash

lean bane
#

open it in notepad i meant

rare gobletBOT
#

Ayo? @lean bane level 1 !!! lfg

wheat rivet
rare gobletBOT
#

Ayo? @wheat rivet level 1 !!! lfg

lean bane
molten pecan
#

Just search "how to make vocals sound brighter"

#

Or how to add saturation

#

Something like that

brittle wing
#

this the the text

lean bane
#

try using mine, but keep a backup of yours obv

molten pecan
#

Don't open it, it's a malicious link

brittle wing
molten pecan
#

Malicious link. Don't open

brittle wing
#

whats in it

lean bane
#

in the manager batch it has an option to fix tensorboard, did you use that? did you use any of the options like update or reinstall?

lean bane
brittle wing
#

i did use the manager batch

brittle wing
lean bane
#

when you open go-applio, what number do you use?

brittle wing
rare gobletBOT
#

Ayo? @brittle wing level 3 !!! lfg

brittle wing
#

i have an amd cpu and gpu so i used 3

lean bane
#

on mine 3 is nvidia with runtime

brittle wing
lean bane
brittle wing
lean bane
#

mine is totally different

brittle wing
#

same

lean bane
#

did you use the install_Applio bat?

brittle wing
#

my file is like this

image.png
lean bane
#

it looks like we have two seperate installations

rare gobletBOT
#

Ayo? @lean bane level 2 !!! lfg

brittle wing
lean bane
#

the whole applio folder?

brittle wing
#

the installations

lean bane
sand iris
#

🍎io moment

#

Quite unstable atm

molten pecan
#

Genuine question

lean bane
sand iris
lean bane
#

any problems i had are for both rvc and applio

lean bane
sand iris
#

So users tend not to pick the newest one

#

And even then the newest one atm is skill issuing

#

...yeah

clever patio
#

Which rvc works normally ffs

#

I ask this alot but none of them work

#

the older versions work no problem

sand iris
#

The "main" RVC project works fine

#

Just a single zip file, extract and profit

clever patio
sand iris
#

pick the latest .7z for your GPU

clever patio
#

say less.

#

ty

plain kraken
#

I have a question

harsh scarabBOT
#

Your request has been added to the queue and will be processed shortly.

plain kraken
#

How do u get a model url

#

In astralab

tame mica
#

💀

plain kraken
#

?

deft reef
#

what exactly model base on?can i load it withtransformer or pytorch and is there document on how to intergrade to my code?

clever patio
#

or do I need just pick 1 7z file ?

verbal tulip
#

Hi, so i have another question if that's ok.
If i want to train a voice model but only have 11 minutes of dataset, what should i set the number for ephochs? 200?

molten pecan
verbal tulip
sand iris
sand iris
#

gl with the code, let's say it's an inter-dependency hell

#

I mean it's python

deft reef
#

oh god

sand iris
#

There's onnx export available, if that's easier to work with

#

But I'm 90% sure the models are like tied to the hubert pertrain they have

slim citrus
#

@sand iris add file not showing what file i should compresse to make zip file nails

image0.jpg
sand iris
#

I don't really know, I'm not doing this on phone

slim citrus
finite galleon
slim citrus
finite galleon
slim citrus
tame mica
slim citrus
tame mica
#

hmm

#

did you stop it without it finishing to the set epoch by chance?

slim citrus
#

Can you tell me other tools like a free website

slim citrus
violet heron
slim citrus
violet heron
#

You don’t get much options but it works decently

rare gobletBOT
#

Ayo? @slim citrus level 3 !!! lfg

slim citrus
#

Ok ok

slim citrus
violet heron
#

You can go lower but it will probably sound eh

#

More is useless and it’s just going to take more time

slim citrus
#

Okay i will make 5 minutes voice

finite galleon
violet heron
#

Thought it was 15

finite galleon
slim citrus
violet heron
slim citrus
sudden tartan
#

I think rvc isnt working

finite galleon
sudden tartan
#

Id send a photo but I cant

#

Basically with every model I use

rare gobletBOT
#

Ayo? @sudden tartan level 1 !!! lfg

stable brook
#

Is Applio currently broken for AMD users?

#

For inference, by the way.

#

Already tried reinstalling for AMD/intel only and tried option [1] for all platforms. CPU works but not AMD GPU.

#

Please ping me when you respond. Thank you.

rare gobletBOT
#

Ayo? @stable brook level 1 !!! lfg

stable brook
#

Output when trying to run Option 3 from go-applio.bat

overwrite preprocess and configs.json
2023-12-05 23:27:19 | INFO | assets.configs.config | Use DirectML instead
Traceback (most recent call last):
  File "F:\Soft\Applio-RVC-Fork\infer-web.py", line 64, in <module>
    import tabs.resources as resources
  File "F:\Soft\Applio-RVC-Fork\tabs\resources.py", line 50, in <module>
    config = Config()
  File "F:\Soft\Applio-RVC-Fork\assets\configs\config.py", line 58, in wrapper
    wrapper.instance = func(*args, **kwargs)
  File "F:\Soft\Applio-RVC-Fork\assets\configs\config.py", line 87, in __init__
    self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config()
  File "F:\Soft\Applio-RVC-Fork\assets\configs\config.py", line 291, in device_config
    import torch_directml
ModuleNotFoundError: No module named 'torch_directml'```
#

Option 1 works but it's only using intel CPU

# 
overwrite preprocess and configs.json
2023-12-05 23:28:43 | INFO | assets.configs.config | Use cpu instead
is_half:False, device:cpu
2023-12-05 23:28:47 | INFO | assets.i18n.i18n | Using Language: en_US
2023-12-05 23:28:47 | INFO | lib.tools.loader_themes | Theme Loaded: Acayun
Running on local URL:  http://0.0.0.0:7897```
stable brook
wind cipher
#

is it normal that my mic stutters when i play a game and talk in voice chat there?

crimson sand
#

Is there any alternative to the tensorboard ROD to detect overtraining?

jaunty shale
#

EasyRVC/Colab

image.png
#

i tried to convert audio

#

with different voice

tame mica
#

easygui?

#

easygui's inference is broken don't use it

glass shoal
#

anybody knows about tortoise tts and how to setup it?

rough oxide
#

So I wanted to install Applio locally on my PC inside a different drive, rather than my local disk since i'm already losing space there, so if I put it in an empty folder inside its corresponding storage drive, will the auto installer install Applio inside that folder?

verbal tulip
#

guys i have a problem with my output audio of mangio rvc, it means the world to me if someone can give me a tip to fix it.z
i used mangio-crepe and experimented with value of hop-length between 1 to 60, i upload the input and output audio so you can see for yourselves

rare gobletBOT
#

Ayo? @verbal tulip level 2 !!! lfg

verbal tulip
spare saffron
#

How do I make it so my model will actually train when I disable pitch guidance?

lethal crown
#

hello, is mango rvc still working? I tried to train a few models many times and they don't sound nothing like the audio I used for the data set. I followed a guide and then everything right. one was from 5 minutes of audio, and the other was 7 minutes. elevenlabs cloned the voice perfect with 5 minutes. but I would rather make my own. but I'm doing everything right on mango rvc it's the newest version of mango rvc v23. 7.0 infer-train on my pc. if anyone

I will go over the steps I take to train.
I make a folder put in mango named it the model name to put the audio in, also name the model the web go when it pops up. under the train tab if I didn't say so before. click 40k click v2 rvc. then I put the path of the folder I named the model name and audio in, where it asks for path of train folder. for the audio. then clicked process data. wait for done. then click rvmpe, then click feature extraction. wait for done. then train feature index. wait for done. then I set my epochs to around 200 because I don't have a lot of audio. set the save to 50 epochs. set the batch size per gpu to 8 to 12, because read the higher it is the better for the training. ( but old pc with a Nvidia gefore gtx 1650 super 4gb and a Dell OptiPlex 9020 i7-4790 3.60GHz 16GB RAM Windows 10 Pro ) then I check two of the boxes and leave the middle unchecked. then train model. then wait over night or how many hours it is.

now you know what steps I do why is the model not sounding like the source audio? the download models work fine in mango rvc. just my trained models don't come out right. the first one I trained was on 5 minutes of audio, at 736 epochs sounded bad so deleted it from the weights folder and the logs folder. and started over around 200 epochs. anyway please help. thank you very much.

mossy fox
#

okay so I trained a model right its 13 minutes long and decent quality for some reason the model trained 400 epochs less then 2 minutes and it doesnt sound bad......what happened?

#

at first it didnt worj

#

anyone have any idea?

slate birch
#

rvc is saying it cant find the zip file in the drive folder, despite the name literally being copied and pasted into the field. Any ideas?

flat harness
#

wahts is the best egirl vocie changer

#

?

sand iris
#

-egirl

azure marshBOT
# sand iris -egirl
Searching for the perfect RVC e-girl voice model?

Look no further – we've got you covered! Our RVC e-girl voicemodel is in high demand, and we've streamlined the process to make it effortlessly accessible for you.

distant turtle
#

https://colab.research.google.com/github/camenduru/tortoise-tts-colab/blob/main/tortoise_tts_mrq_colab.ipynb?pli=1&authuser=0 I'm using the colab version of Tortoise - tts (because my computer is too old and consequently I can't complete the installation), do you have a more updated link?, I tried to train a model but I have 2 problems, one not I think there is the Italian language because in the model I created I wrote in the language: IT, then when I generate the text it either gives me errors or gives me a terrible result, do you have an updated link?

hoary magnet
#

Hi um ..
I need help its shows pipeline is not initialized

molten pecan
half cove
distant turtle
# half cove your trying to train on tortoise?

no, I have already created a model, but even if I put the Italian language, the model opens very badly in Italian, and is not at all similar to the voice I gave, I put 3 minutes of very clean audio, maybe there is a more updated site, because it easily fills up the free memory and as a result I can no longer generate speech from text

half cove
#

so you have trained on tortoise?

distant turtle
lethal crown
half cove
molten pecan
#

How long is your dataset?

lethal crown
#

5 minutes of audio and 200 epochs

rare gobletBOT
#

Ayo? @lethal crown level 1 !!! lfg

distant turtle
unkempt halo
#

Anyone got any tutorials?

molten pecan
#

Like 700

molten pecan
unkempt halo
lethal crown
#

I did try 736 epochs the first time. but it didn't sound good either, but I will try again. thank you. but think something else is not going right.

#

so I dropped the epochs to 200 after

molten pecan
#

Because, like the person above said, for short dataset use higher epochs

#

And for longer ones use shorter epochs

lethal crown
#

I'm feeling hopeful now 👍

molten pecan
lethal crown
molten pecan
#

I'm not familiar with making models so give it a shot if you want

lethal crown
rare gobletBOT
#

Ayo? @boreal mango level 2 !!! lfg

prime hemlock
#

I AM DOING A RVC HELP DESK IF ANYONE NEEDS HELP

astral perch
#

What does the batch size do?

ripe hatch
#

hi guys

#

wheres the link to voice clone

#

im new to this hah

#

i alrrayd have the model i just wanna do ai stuff

molten pecan
azure marshBOT
#

Suggestions for @ripe hatch

All Working Local Forks 🖥️
Need some help? 🤔

You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.

More commands: -audio, -colabs, -overtrain, -realtime, -rvc, -help
ripe hatch
#

hello

#

can anyone help me?

#

why does my model not show up in "1. choose the model" in illaria rvc? it says it downloaded

brittle wing
#

hit refresh

#

then select it

#

anyone know a good way to isolate vocals

molten pecan
molten pecan
#

I'm the main tab

limber nimbus
#

Why I can't merge theese model?
Model1

采样率:40k
模型是否输入音高引导:1
版本:v2```
Model 2
```模型信息:1000epoch
采样率:48k
模型是否输入音高引导:1
版本:v2```
Result
```Traceback (most recent call last):
  File "E:\DATA\AI STUFF\RVC\Mangio-RVC-v23.7.0\train\process_ckpt.py", line 242, in merge
    alpha1 * (ckpt1[key].float()) + (1 - alpha1) * (ckpt2[key].float())
RuntimeError: The size of tensor a (24) must match the size of tensor b (16) at non-singleton dimension 2```
exotic adder
#

would anyone be willing to kindly help me find an ai filthy frank voice so that i can make something

violet heron
exotic adder
#

it must. ive seen youtube videos with the ai voice of filthy frank

viscid moss
obtuse stream
#

anyone have Applio-RVC-Fork-main.zip
not the latest

dawn nebula
rare gobletBOT
#

Ayo? @dawn nebula level 3 !!! lfg

brittle wing
pallid heron
#

what does crepe hop do + feature retrieval rate

jaunty iris
jaunty iris
#

feature ratio is for how much of the models index is used

#

the index carries the accent of the voice it was trained on

pallid heron
hexed dew
jaunty iris
pallid heron
rare gobletBOT
#

Ayo? @pallid heron level 1 !!! lfg

pallid heron
#

between crepe, crepe tiny, harvest, dio, pm

finite galleon
#

But there are many methods that are better than crepe like mangio-crepe, rmvpe, and fcpe

pallid heron
#

gotchu

pallid heron
finite galleon
jaunty iris
finite galleon
jaunty iris
#

its a bit annoying but worth it

thin stump
hexed dew
thin stump
#

I don't know how to remove harmonies or choirs. I'm hoping that's an option down the line

brittle wing
#

Can I please ask someone to help me

#

To make a cover and remove the breathiness

hexed dew
finite galleon
#

Y u ping

#

For the record i dont use local web uis

#

But litteraly applio is a pain to install from what i have sern

distant turtle
#

https://colab.research.google.com/github/camenduru/tortoise-tts-colab/blob/main/tortoise_tts_mrq_colab.ipynb?pli=1&authuser=0 I'm using the colab version of Tortoise - tts (because my computer is too old and consequently I can't complete the installation), do you have a more updated link?, I tried to train a model but I have 2 problems, one not I think there is the Italian language because in the model I created I wrote in the language: IT, then when I generate the text it either gives me errors or gives me a terrible result, do you have an updated link?

sand iris
#

fairly sure blaise is aware of the python dependency issues by now

peak tusk
#

The installation works correctly in most cases but for some reason on some computers it fails and there are different parts of the process where that can happen, so we need you to open an issue specifying the hardware and everything possible on your computer to try to replicate it

sand iris
#

At least with the ipex stuff

#

tensorboard was like expected to be older by some libs, when others wanted only newer versions

obtuse stream
#

anyone have Applio-RVC-Fork-main.zip
not the latest like this Applio-RVC-Fork-main.zip

edgy nova
#

what matters more, the amount of epochs or the amount of steps?

#

the model i'm training atm is at 120 epochs and 7080 steps so far

#

sorry if that's a stupid question, i've never paid attention to the steps before until now

proper shale
#

More steps per epoch means more training

#

But not necessarily a better model because of overtraining

edgy nova
#

alright

#

how's this?

image.png
distant turtle
#

https://colab.research.google.com/github/camenduru/tortoise-tts-colab/blob/main/tortoise_tts_mrq_colab.ipynb?pli=1&authuser=0 I'm using the colab version of Tortoise - tts (because my computer is too old and consequently I can't complete the installation), do you have a more updated link?, I tried to train a model but I have 2 problems, one not I think there is the Italian language because in the model I created I wrote in the language: IT, then when I generate the text it either gives me errors or gives me a terrible result, do you have an updated link?

upbeat talon
#

Hello everyone, my Voice real Time Voice changer does not work, namely, the changed sound is not output? what is wrong?
NVIDIA GeforceGTX 1060 6 GB video card

upbeat talon
#

settings like this

proper shale
proper shale
#

crepe is way too slow for realtime

upbeat talon
edgy nova
#

i'll just keep that running until colab kicks me out

proper shale
#

uhhhh

#

are you on RVC disconnected

#

if so you can already stop and export model

edgy nova
#

alright

upbeat talon
proper shale
#

what does cmd say

upbeat talon
#

Maybe it's because I didn't download the virtual cable?

proper shale
#

nah I don't think it is

#

just check the command tab

upbeat talon
proper shale
#

yup, not your GPU's fault nor your f0

the initial installation went wrong

#

try this:

delete pretrain and model_dir and restart W-Okada

upbeat talon
proper shale
#

should reinstall all the models and pretrains

proper shale
# upbeat talon

if in case one of the installs doesn't work, it throws that message

upbeat talon
proper shale
#

it just does that once you delete pretrain and model_dir

#

and start w-okada

upbeat talon
proper shale
#

Delete the two folders and restart W-Okada

upbeat talon
warped mural
#

guys how do i fix when im using w okada and i press start in the command prommt it says pipeline not installed and generating pipeline and again not installed

proper shale
#

btw if this is copyrighted pls remove it

#

but like it sounds alright

edgy nova
#

oh shit

edgy nova
proper shale
warped mural
proper shale
#

delete em both then

warped mural
#

i think i got this error beacus when i was installing w okada i didnt have python if your method works thank you

rare gobletBOT
#

Ayo? @warped mural level 1 !!! lfg

upbeat talon
proper shale
proper shale
upbeat talon
upbeat talon
proper shale
#

uhhh

#

how can i say this

#

the folder Vice mod has a space on it, that could be causing problems

#

also delete the stored_setting.json file and restart W-Okada

upbeat talon
upbeat talon
#

And him crashed

#

But google colab working

warped mural
upbeat talon
upbeat talon
glad zealot
warped mural
glad zealot
#

use it

harsh scarabBOT
#

You're placed in Cover Generator Queue. Current position: 1

brittle wing
#

HELP

#

MY RES BE HIGH AS HELL HOW I GET IT LOWER

crimson sand
#

Hey, can someone confirm this training loss graph looks healthy? (0.99 smoothing, hide outliers off)

slate pawn
#

any can help with this? The index is not working for a certain model despite training it

cmd_9pWKcbUXto.png
eternal fern
#

hey, i have some questions about how to making a high quality path with RVC (i know how to make. just need to know what should be changed)

proper shale
#

Ask away!

#

I'll try my best to answer them

cobalt oyster
#

Can anybody say me best photo upscaler colab could anybody give me link pls

proper shale
#

use this

fossil glen
molten pecan
molten pecan
molten pecan
rare gobletBOT
#

Ayo? @limber trout level 1 !!! lfg

limber trout
molten pecan
#

Not selecting it from the list

viscid moss
karmic bone
#

guys, i accidentally deleted something in my Google Drive when i was cleaning it

#

and then

#

the RVC stop worked

#

it shows me TensorBoard with error

#

i don't know how to fix it, now i am sad((

viscid moss
karmic bone
#

doesn't anyone know what to do? ((

viscid moss
karmic bone
#

what is it?

#

i deleted everything except soke zips files

rare gobletBOT
#

Ayo? @karmic bone level 1 !!! lfg

karmic bone
viscid moss
karmic bone
#

damn....

#

i feel so stupid now

#

can i download it again somehow?😭

viscid moss
#

You will have to retrain the model bro

viscid moss
karmic bone
#

i download another model

#

and still not works

#

or i need to delete the old zips too and make new ones?

viscid moss
#

send me a screenshot

karmic bone
#

ok

#

i am just scared i never can return it((

viscid moss
#

-colab

azure marshBOT
jaunty shale
#

any recommendation for UVR model that will extract vocals perfectly?

image.png
viscid moss
proud elbow
limber trout
teal creek
#

Which RVC is better W-Okada or Mangio-RVC?

molten pecan
#

Okada is for real-time voice conversion. Like, for calls

#

And Mangio is for inference and training models

#

Like AI covers

teal creek
#

I see

#

And it requires RTX series?

jaunty shale
molten pecan
#

Even from your phone

teal creek
#

Got it, thanks a lot.

molten pecan
weak bloom
#

helpClick on "Download Models" and input the URL of the desired model from Forum AI Voice Models and Model Link Sites like Weights.gg. (IF THE URL IS https://weights.gg/download?fileUrl=... OR https://models.weights.gg/…

dreamy schooner
#

help

#

Error: 'f' failed: could not open models\BarısAkarsu5555\added_IVF241_Flat_nprobe_1_BarìsAkarsu_v2.index for reading: No such file or directory

weak bloom
molten pecan
proud elbow
odd shale
#

@grim rapids C'mon buddy, ask for help right here.

#

@viscid moss Tengo una amiga gringa que quiere instalarse el W okada.

proud elbow
odd shale
grim rapids
odd shale
grim rapids
jaunty shale
proper shale
#

uh

proud elbow
# jaunty shale

install python 3.9.8 first, or try copy python39.dll from the python dir to the UVR dir?

proper shale
#

remove the space

#

in the folder name

molten pecan
proud elbow
#

oh yeah no spaces

proper shale
#

in both folder names actually 💀

#

both rvc stuff and uvr

static tapir
#

How am I going to make a voice model?

jaunty shale
#

oh wait

proper shale
#

hope these will help

jaunty shale
odd shale
jaunty shale
#

@proud elbow what is the most recommended settings before processing?

odd shale
#

Hi little buddy

proud elbow
granite bluff
#

-colab

azure marshBOT
static tapir
proper shale
#

In Google Drive? just configure the share options so that everyone wth the link can see

slate pawn
#

can anyone help me with this index problem

cmd_9pWKcbUXto.png
brittle wing
#

anyone know how to fix this?

image.png
shy turret
proper shale
#

to work anywhere outside the app, yeah

brittle wing
shy turret
#

it was so crazy frustrating for me. I followed all of the steps making certain devices default and others left ready but there is just no output . The input works but no output. I cant use the output for vb its dead

brittle wing
sand iris
#

Unfortunately that's just Applio having issues at the moment

#

Not sure what happened, but the python packages are all fucked

solemn walrus
#

is this where i get helpf or the voice changer?

sand iris
crimson sand
#

Hey is it normal that Applio only updates the modelname\eval tensorboard log?

night stone
#

Is there any link to a guide on how to create an RVC-V2 Model using my own hardware, I dont want to use one of those collab spaces

molten pecan
static gyro
#

how to get a file and download it please?

image.png
viscid moss
static gyro
rare gobletBOT
#

Ayo? @static gyro level 11 !!! lfg

brittle wing
#

how can i find the masculine and feminine out there?

sweet cradle
#

I tried using Mangio rvc instead of Applio. When trying to use audio files for the dataset, clicking the Process Data button throws a value error.

Traceback (most recent call last):
  File "C:\Users\Alexian PC\Desktop\Mangio-RVC-v23.7.0\trainset_preprocess_pipeline_print.py", line 8, in <module>
    sr = int(sys.argv[2])
ValueError: invalid literal for int() with base 10: 'PC\\Desktop\\Mangio-RVC-v23.7.0\\datasets\\'```
sand iris
#

Your username has a space in it's name

#

All the RVC stuff absolutely dies when there's any spaces in the file paths lol

static gyro
# static gyro

In summary, the log provides information about the configuration, loading of pre-trained models, and errors encountered during the training process, specifically related to loading checkpoint files and a size mismatch in model parameters.

#

what checkpoint am i loading here?

#

im making a new model

#

it hasnt made it yet

#

im confused

distant turtle
#

why this error?, can anyone give me a solution, Thanks

Screenshot_5.png
brittle wing
#

why level?

sweet cradle
rare gobletBOT
#

Ayo? @sweet cradle level 2 !!! lfg

sand iris
#

Unsure, there are issues sometimes but give it a shot

sweet cradle
#

Looks like the error is gone. Thanks a bunch. :)

brittle wing
#

what are the best things to set in rvc2? like to get the best quality output aside from using uvr to clean the acapella thats being covered

molten pecan
#

If you try mangio-crepe lower the hop length until you find a sweet spot

#

By default set search feature ratio to 90

#

If there is some artifacting during breathing or exhalation, lower the "protect voiceless consonants" value

brittle wing
molten pecan
#

If there is artifacting overall, lower the search feature ratio and/or use rmvpe

#

You'll find out yourself what the best settings are when you try

brittle wing
#

i usually use rmvpe but it dosnt come out the best most of the time, sometimes crepe does but everything just sounds very hissy

molten pecan
# brittle wing what does that do to it?

The time it takes for the voice to reach a certain note. The lower, the more accurate it will sound to the original audio. But setting it too low sounds a bit off, in my experience

#

10-150 is usually where it sits best

brittle wing
#

fuck now some of my shitty covers can sound good lol

molten pecan
molten pecan
#

It's all trial and error. Have fun and see it for yourself

slate birch
#

hwo do I make an ai model do TTS?

brittle wing
#

usually i have to attempt to fix the sound with post processing 😭 but i will say using locally works far better than when i was using collab, even before google took it

molten pecan
molten pecan
brittle wing
#

this is what its usually at

image.png
molten pecan
slate birch
molten pecan
#

But that's not a problem

slate birch
#

oh?

molten pecan
#

Run the text through the TTS and then convert the output with your model

brittle wing
#

audios i just do itunes loopback and then uvr to seperate the vocal bc the m4a from itunes has been best quality compare to other m4a files ive seen and tried

rare gobletBOT
#

Ayo? @brittle wing level 2 !!! lfg

molten pecan
#

You can still do what you want to do, it only requires one extra step

brittle wing
#

but im more into the post processing ive made some rly cool sounding covers now compared to my first few😭

brittle wing
#

where do i adjust the hop?

molten pecan
#

Uhm

#

You do know that mp3 format compresses the audio quality, right?

brittle wing
#

yea

molten pecan
#

It's best to have good quality audios, preferably in lossless quality

brittle wing
#

i do mp3 for like backing vocals

molten pecan
#

So RVC works best

brittle wing
#

alot of the covers i do have distortion added, so i dont rly try hard with the backing compression

#

i do the lead vocal in lossless and the instrumental seperately

molten pecan
#

So I'm just saying, if what you want is good results in RVC, avoid audio files with lossy quality

molten pecan
#

The hop length setting should pop up

brittle wing
#

oki thxx

molten pecan
#

It should in theory

slate birch
#

does anyone have advice on making a good ataset of my own voice? I made one of me talking for about half an hour but it sounds crackly and not like me

brittle wing
#

what does median filtering do?

#

like what should i set it at

#

and search should it be 0.90 or just 90

molten pecan
#

So don't worry about kt

#

It*

brittle wing
#

ok

molten pecan
brittle wing
#

how to i make it sound less breathy? like it sounds breathy/metalic

molten pecan
brittle wing
#

what do u think a good value for it is

molten pecan
#

To like, 0.25

brittle wing
#

oki lemme try

molten pecan
#

@brittle wing also remember, you can make the output sound lower or higher naturally

#

If you want it to

brittle wing
#

wdym

molten pecan
#

In "Transpose"

brittle wing
#

like pitch

molten pecan
#

You can change the pitch

#

Yes

brittle wing
#

ik that thats the only thing i ever changed 😭

molten pecan
#

You can use decimals if you wish. Like -1.5 or 0.6 for example

#

Alright

brittle wing
#

i didnt wna break the other shit lol

#

the 0.25 did help

molten pecan
#

Awesome

brittle wing
#

now it sounds less like shes blowing in the mic

molten pecan
#

Cool

#

Are you even going to say thank you for all the help 😭

brittle wing
#

i am im testinggg 😭

#

omg lol

#

thnk u this has helped alot, now i feel btr ab my covers n dont have to do as much filtering 😭

molten pecan
#

Yw

#

Good luck with everything

brittle wing
#

soo what if my other model sounds rly squeeky on the same one a different model was fine with, when they are both mangio-crepe models

rare gobletBOT
#

Ayo? @brittle wing level 3 !!! lfg

molten pecan
#

Every model sounds based on its dataset

#

If you trained the model using audios where the voice sounds squeaky, then the model will sound squeaky

brittle wing
#

ok, apparently i pressed the wrong thing 😭 thats why it sounded bad

#

bc i redid it and its fine now

real fable
#

how do i download the cover from the colab version of covergen

violet heron
brittle wing
#

is there a way to train online again? i wna update my model but dont have enough vram

violet heron
azure marshBOT
# violet heron -colab RVC Disconnected

Suggestion for @brittle wing

All working colabs / spaces ☁️

Credits to Faze Masta and Antasma for compiling these links.

real fable
#

refresh collab

real fable
violet heron
real fable
molten pecan
#

@gentle helm did you duplicate the space?

gentle helm
#

no, i did not .....

#

i forgot to do that

molten pecan
#

Ping me if anything!

gentle helm
#

TYSMM

molten pecan
#

Yw!!!!

gentle helm
#

okok, ill lyk how it goes!

molten pecan
brittle wing
#

how do i add my files to the huggingface space?

#

i have my dataset made i just dont know what 2 do next

half cove
molten pecan
#

:3

violet heron
#

You have to use RVC Disconnected

gentle helm
molten pecan
#

Just ignore it

half cove
#

means the model doesnt have index

#

its for

#

accents

molten pecan
#

It's really weird 😭

half cove
#

which space?

molten pecan
#

But don't worry it still acts as if it detected the index

gentle helm
#

and for the inferencing voice, the only thing that shows up is logs/weights/model.pth

#

is that right?

molten pecan
gentle helm
brittle wing
#

heyy

rare gobletBOT
#

Ayo? @brittle wing level 1 !!! lfg

brittle wing
#

can anyone help me? i'm having trouble isolating vocals on uvr5

molten pecan
molten pecan
violet heron
brittle wing
violet heron
#

If not try X-minus pro

brittle wing
#

which one's best?

violet heron
brittle wing
#

mmm i see

#

i'll give it a few more tries

#

thx

half cove
burnt breach
#

Hello, newbie here, i would like some help with few things.
I wanted to do some AI covers for fun and i started to follow a github guide and all by SociallyIneptWeeb. But i'm facing few problems, first of all, my python ver is 3.10 instead of 3.9 like it's recommended so i don't know if it will be troublesome at some point. Also, i'm at a moment where he ask :

Download required models
Run the following command to download the required MDXNET vocal separation models and hubert base model.

python src/download_models.py

#

but i get somewhat an error in my commands

#

I'm not sure anymore if i should push all that, i feel very lost : (

sand iris
#

What's the error you are getting ?

molten pecan
half cove
molten pecan
half cove
#

zip

molten pecan
#

That's weird then 😭

#

Prob you pasted the wrong link

half cove
burnt breach
rare gobletBOT
#

Ayo? @burnt breach level 1 !!! lfg

sand iris
#

Oh yeah, you have to do that from the correct directory

molten pecan
#

So they don't have to go through the whole process of uploading it to HF

half cove
#

it uses applio 😡

sand iris
#

I'm too tired atm, I'm sure someone here will be able to point you in the right direction

molten pecan
#

Dang

#

That explains your bug real

burnt breach
half cove
sand iris
molten pecan
#

Is Ila RVC built from scratch?

half cove
#

they got my boy vidal doing 18 hour shifts 😭

image.png
half cove
burnt breach
sand iris
#

It's where you unzipped/downloaded

#

Not sure what instructions you are following lol

candid jetty
#

anyone know what could be causing this

image.png
#

(I have a nvidia gpu)

sand iris
#

a) no nvidia gpu
b) missing nvidia driver (for the CUDA stuff)
c) wrong torch install

#

how'd you set it up ?

candid jetty
sand iris
#

how'd you install torch specifically ?

candid jetty
burnt breach
candid jetty
#

is it a different version of CUDA?

sand iris
#

I think either work

#

Just make sure you had windows selected

candid jetty
#

I did lol

sand iris
#

else you'd be installing the "stock" torch which is CPU only

candid jetty
#

lemme try again

sand iris
#

pip3 install torch torchaudio --index-url https://download.pytorch.org/whl/cu121

#

without the extra url it fetches the CPU only build

candid jetty
#

kk

#

lemme restart conda env rq

unborn oasis
#

extract_f0_festure.log error

I asked chatgot about this error that occurs to me everytim I try and train a voice model.

I can't find the file anywhere. Apparently, it's because it doesn't exist. It's supposed to be like a stamp log that keeps track. If it doesn't exist, it's simply created. It's a Unix thing, so the directory file paths syntaxes can be different. If there is spaces in a folder name, it does not read the lath correctly.

Apparently, all I would need to do is copy my file paths in quotes to solve this.

Anyone else have this issue?

unborn oasis
gentle helm
molten pecan
gentle helm
#

that’s the only one that shows up

IMG_6621.jpg
#

is that the right one?

molten pecan
#

Yes

molten pecan
#

Keep following the guide

gentle helm
#

it’s not showing up in the one on the right

rare gobletBOT
#

Ayo? @gentle helm level 2 !!! lfg

molten pecan
#

Did you make yourself that model?

gentle helm
gentle helm
molten pecan
#

Can you still convert it eitherway?

#

Does it allow you?

gentle helm
#

it’s still showing the error message

candid jetty
molten pecan
#

Or are the numbers still running?

gentle helm
sand iris
candid jetty
#

I dont think so

molten pecan
# gentle helm

Click the three dots, go to Settings, and scroll until you find Restart this space

#

That might fix it

sand iris
#

Might be missing the CUDA toolkit driver from nvidia then

#

Really don't know what's up with it, some users just don't have to install it

candid jetty
gentle helm
#

thank you smm! 😭

sand iris
#

yeah

molten pecan
ancient cloak
#

How do you get better quality with covers?

rare gobletBOT
#

Ayo? @ancient cloak level 3 !!! lfg

violet heron
finite galleon
ancient cloak
#

I have both already

#

But are there ways to increase quality in like uvr 5 or something? Also whats the best program to use for doing the actual model to song conversion

molten pecan
#

Or Windows Size. I don't remember the name

#

I think the maximum value available was 1024

#

Also make sure you use an audio in lossless quality and format. Avoid mp3s

candid jetty
sand iris
candid jetty
#

idk if u know Rxhul but me and him have been trying for like a full 2 days lol

#

I have no clue at this point

sand iris
#

what does this python script output ?

import torch

print(torch.cuda.is_available())
#

also IIRC when running pip to manage packages, it can just not update stuff, so it might've kept CPU only builds ot torch ?

candid jetty
sand iris
#

yeah you gotta run it within python

candid jetty
#

should it not work within a python environment?

gentle helm
#

is there any way i can change the name of the ai models i import?

sand iris
candid jetty
sand iris
#

all the venv does is change the default python interpreter, the python packages, and where to find external software, to something that's not system wide

#

conda supplies python

#

if you just python, you'll get the python command line instead

candid jetty
sand iris
#

was the virtual env activated on that terminal ?

candid jetty
sand iris
#

Try RVC again trol

#

It might just work now

candid jetty
#

😭

#

I swear if this works

ancient cloak
candid jetty
sand iris
#

what

candid jetty
sand iris
#

What's your GPU again

candid jetty
#

4070

sand iris
#

Shit's cursed somehow

candid jetty
#

I dont get ittt

#

cuda and drivers are definitely fine

image.png
sand iris
#

Also is there any reason you are going the venv way with the RVC releases ?

candid jetty
#

wdym

#

how else can I do it?

sand iris
#

The RVC100Nvidia archives available on the RVC-webui hugging face page

#

they bundle python and everything else RVC needs

candid jetty
sand iris
#

Yup

candid jetty
sand iris
#

Ah lol

candid jetty
sand iris
#

You're supposed to go the venv way with the bare git repository (even though it should still work)

sand iris
#

Wait it's not extracted?

#

Might be why

proud elbow
rare gobletBOT
#

Ayo? @proud elbow level 10 !!! lfg

candid jetty
sand iris
#

Oh yeah lol I wouldn't make any sense it if wasn't kittyblep

candid jetty
#

wait why is that a wav lol

#

wait a sec

candid jetty
sand iris
#

don't run as admin !

candid jetty
sand iris
#

what does it do when executed normally ? Instantly close?

candid jetty
#

just sits like this

sand iris
#

Now that's cursed

#

Maybe try letting it be for a while, but I doubt it'll do much

#

really not sure what to do from here

candid jetty
#

I'll try get a response from kalo ig lol

#

but rxhul knows rvc rlly rlly well and didnt know what to do so im not too optimistic tbh

candid jetty
#

no way 😭

image.png
#

thats funny as shit

violet heron
candid jetty
#

I like opera

violet heron
#

https://rentry.co/operagx

No Opera

Opera (GX/One/Crypto Browser) is a closed-source browser owned by a company with a track-record of selling user-data, censorship, backdoored software and involvement in predatory loans and anti-competitive behavior. While the link for this page is "OperaGX", this applies to every browse...

candid jetty
#

they can have my data its chill

violet heron
candid jetty
#

as long as they give me workspaces im happy

#

and Aria is cool

#

saves me from keeping chatgpt open

candid jetty
digital pulsar
#

ok how do i do it

#

how do i put downloaded voice in here

#

alright no image perms

#

how to put voice into Inferencing voice:

finite galleon
digital pulsar
#

helpppp

violet heron
#

Press refresh

slate birch
#

On the AICoverGen web UI, i’ve been running self made models and downloaded ones as well to make covers of music, just experimenting. They all sound kind of crappy though. I think it has something to do with the “ConnectionResetError” that happens at the final step of the ai cover generation. Does anyone know anything about that or have any ideas as to how to fix it?

rare gobletBOT
#

Ayo? @slate birch level 2 !!! lfg

latent kettle
latent kettle
violet heron
#

RVMPE > Mangio-Crepe > FCPE > Harvest > Crepe > Crepe-tiny > PM > DIO

latent kettle
finite galleon
latent kettle
viscid moss
shy karma
#

Hi guys, what is the best current local rvc?

rare gobletBOT
#

Ayo? @shy karma level 1 !!! lfg

glad zealot
copper pebble
#

mornin chat

#

even tho i dunno what time is it for ya

proper shale
proper shale
shy karma
#

@proper shale thank you. What is the mainline? Or is it just the name of it?

copper pebble
#

wait why dont you go to sleep bro

proper shale
shy karma
#

thank you

quaint nacelle
#

Traceback (most recent call last):
File "C:\Applio-RVC-Fork\infer-web.py", line 55, in <module>
from lib.infer.modules.vc.modules import VC
File "C:\Applio-RVC-Fork\lib\infer\modules\vc\modules.py", line 12, in <module>
from lib.infer.infer_libs.audio import load_audio
File "C:\Applio-RVC-Fork\lib\infer\infer_libs\audio.py", line 3, in <module>
import av
File "C:\Users\aylin\AppData\Local\Programs\Python\Python39\lib\site-packages\av_init_.py", line 46, in <module>
from av._core import time_base, library_versions
ImportError: DLL load failed while importing _core: The specified module could not be found.
Press Enter to continue...

Please help appolio

shy karma
#

hellp guys, I am having an error when trying to install mangio's requirements. could you help me?

#

note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

#

nvm it is about python version i see now

red violet
#

try clicking everything but meh. stuck on sign in options page

robust quartz
#

what is connet sid?

sand iris
#

nothing relevant for you

glad zealot
glad zealot
sand iris
#

nah I think they're having issues

glad zealot
#

nvm its actually bugged

#

anyways heres the link for that model

#

they patched the method im using to download models without logging in lol

#

found another way

#

@sand iris can ya help try making a script for this bypass?

sand iris
#

sure ig

glad zealot
#

dm?

sand iris
#

👍

crimson sand
#

If I have a primarily 20khz (target 40khz) dataset with one larger sample that's 16khz (so target 32khz) is that okay? or should I just throw out the 32khz sample

slate birch
proper shale
#

I think Eddy got confused and thought you were running on Colab

proper shale
astral jungle
#

Yooo

bleak locust
#

i cant use the crepe

grizzled blaze
#

Is the tensorboard included with applio rvc?

#

It seems i cant use it

#

do i have to download it or smth?

crimson sand
#

you have to run go-tensorboard.bat

grizzled blaze
crimson sand
#

HOWEVER... Applio doesn't update tensorboard properly for some reason

grizzled blaze
#

i got into and the cmd got instantly closed

crimson sand
#

oh are you in powershell

#

you have to run it in cmd, otherwise it just closes

grizzled blaze
#

but im gonna try that anyway

#

how can I open it with the cmd

#

any command?

crimson sand
#

just make sure youre not launching powershell, the issue iirc is to do with the python venv stuff

grizzled blaze
#

im not

#

i havent ever used powershell

crimson sand
#

okay uh, try running the commands in the .bat file manually

rare gobletBOT
#

Ayo? @crimson sand level 2 !!! lfg

crimson sand
#

make sure you're in the right directory

#

iirc it's a PATH issue with the venv and the script closes before it can show you the error

grizzled blaze
#

how can i solve it

crimson sand
#

So cd to the applio directory

#

then run cmd.exe

# 
set CONDA_ROOT_PREFIX=%UserProfile%\Miniconda3
set INSTALL_ENV_DIR=%cd%\env

echo Activating Conda environment...
"%CONDA_ROOT_PREFIX%\condabin\conda.bat" activate "%INSTALL_ENV_DIR%"

if errorlevel 1 (
    echo Failed to activate Conda environment.
    pause
    exit /b 1
)

echo Starting the Tensorboard with Conda...
cls
echo Starting Tensorboard...
python lib\fixes\tensor-launch.py
#

Then paste that

grizzled blaze
#

cd?

#

copy directory?

crimson sand
#

no the comman cd

grizzled blaze
#

oh lmao

crimson sand
#

okay wait

#

press windows+r

#

and type cmd.exe

#

then type cd C:\Wherever\Applio\Is

sand iris
#

applio moment

crimson sand
#

actually wait, easier way to do this

grizzled blaze
#

how do i change the disk

crimson sand
#

open up go-tensorboard.bat, and delete the exit at the end

#

and run it

#

it should flash the error but stay on screen now, IIRC it's the miniconda installation

grizzled blaze
crimson sand
#

yeah so

#

at the end of go-tensorboard there's exit /b 0

#

just remove that line

grizzled blaze
#

ohhh

#

ok ok

crimson sand
#

now run the bat, and screenshot the error

grizzled blaze
#

it still closes

crimson sand
#

it shouldn't... wat

#

the pause should keep it open

#

okay, where is your applio installed

#

give me the full path

grizzled blaze
#

X:\Games\RVC\Applio-RVC-Fork

crimson sand
#

ok

grizzled blaze
#

ignore why is the file called games

crimson sand
#

Windows Key + R
cmd.exe
cd X:\Games\RVC\Applio-RVC-Fork
set CONDA_ROOT_PREFIX=%UserProfile%\Miniconda3
set INSTALL_ENV_DIR=%cd%\env
echo %CONDA_ROOT_PREFIX%
echo %INSTALL_ENV_DIR%

#

then show me the output

grizzled blaze
#

nothing happens

image.png
rare gobletBOT
#

Ayo? @grizzled blaze level 3 !!! lfg

crimson sand
#

yeah it wont until you run echo

grizzled blaze
#

oh ok

crimson sand
#

set doesnt output anything, it just sets the local env

grizzled blaze
#

nothing happened

image.png
crimson sand
#

no something did happen

#

thats good

#

"%CONDA_ROOT_PREFIX%\condabin\conda.bat" activate "%INSTALL_ENV_DIR%"

#

now run that

grizzled blaze
#

with the " and the %?

crimson sand
#

yes

#

copy it exactly

grizzled blaze
crimson sand
#

okay

#

applio didn't set up your miniconda properly

#

one sec

#

"%CONDA_ROOT_PREFIX%\_conda.exe" create --no-shortcuts -y -k --prefix "%INSTALL_ENV_DIR%" python=3.9

#

then run

#

"%CONDA_ROOT_PREFIX%\condabin\conda.bat" activate "%INSTALL_ENV_DIR%"

grizzled blaze
#

k

crimson sand
#

it should change to say (env) before C:\Users\Admin

grizzled blaze
#

"System couldnt find the specified route"

image.png
crimson sand
#

ok one sec

#

run dir %CONDA_ROOT_PREFIX%

grizzled blaze
#

the volume of the unit C doesnt have a label
The serie number of the volume is:

Cant find the file

image.png
crimson sand
#

wat

rare gobletBOT
#

Ayo? @crimson sand level 3 !!! lfg

crimson sand
#

ok.. try

#

"%CONDA_ROOT_PREFIX%\condabin\conda.bat create --no-shortcuts -y -k --prefix "C:\Users\Admin\env" python=3.9

grizzled blaze
#

"The name of the file, name of the directory or the sintaxis of the label of the volume are not correct

image.png
crimson sand
#

are you running this all in the same terminal?

grizzled blaze
#

yes

crimson sand
# grizzled blaze

because this clearly shows you have %CONDA_ROOT_PREFIX%\condabin\conda.bat

#

okay.. uh

#

"%UserProfile%\Miniconda3\condabin\conda.bat create --no-shortcuts -y -k --prefix "C:\Users\Admin\env" python=3.9

#

run that

grizzled blaze
#

same message

crimson sand
#

my brother in christ

#

ok wait there

viscid moss
viscid moss
crimson sand
#

@grizzled blaze wait

#

did you run the installer bat?

#

or did you just clone it

grizzled blaze
#

applio installer?

#

ye

crimson sand
#

yeah

#

okay so if you go to windows file explorer

#

and go to C:\Users\Admin\Miniconda3

#

what is there

grizzled blaze
crimson sand
#

what the fuck

#

okay

#

"C:\Users\Admin\Miniconda3\condabin\conda.bat create --no-shortcuts -y -k --prefix "C:\Users\Admin\env" python=3.9

#

run that

grizzled blaze
#

same message

crimson sand
#

bro what

#

is there a conda.bat in the condabin folder?

#

OH

#

WAIT

#

C:\Users\Admin\Miniconda3\condabin\conda.bat create --no-shortcuts -y -k --prefix "C:\Users\Admin\env" python=3.9

#

run that

#

fucking quotes

grizzled blaze
#

it answered