#✨│ai-help
soft stratus
low shard
You are using the realtime voice changer?
Or the rvc tool for making ai covers
I saw alow of people having that error abt wokada so i thought it was local wokada, if im wrong sorru
low shard
https://colab.research.google.com/drive/1ECT_lkB6f0ljpSDd6xTsQOZfYrotEvBE?usp=sh...
Where did you even get that colab
soft stratus
Where did you even get that colab
Since 4 months
dark wing
Hello everyone. I forgot to check the "keep only newest cpth file to conserve disk space" when training a model.
Which .pth files are safe to delete in this case?
low shard
Since 4 months
Thats for sure outdated, check this https://docs.aihub.wtf/essentials/how-to-make-ai-cover/
soft stratus
Thats for sure outdated, check this https://docs.aihub.wtf/essentials/how-to-mak...
I didn't find a link to get me to RVC
low shard
I didn't find a link to get me to RVC
Check on the cloud
There will be colabs
You have to read and see the pros and cons of them based on your needs
dark wing
Hello everyone. I forgot to check the "keep only newest cpth file to conserve di...
do I only keep the two pth files with the largest number?
I see that the number of pth file pairs corresponds to the number of epochs
soft stratus
But i want easy gui
dark wing
Hello everyone. I forgot to check the "keep only newest cpth file to conserve di...
Wait, all the .pth files are safe to delete?
rare gobletBOT
Ayo? @dark wing level 1 !!! 
low shard
But i want easy gui
This is the new easygui https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scrollTo=Sb5fzhzEXK8X
azure marshBOT
RVC Guides (How to Make AI Cover)
soft stratus
This is the new easygui https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0Z...
But it says when i click the download model error
low shard
But it says when i click the download model error
Click the download model? You used an outdated colab, thats why, you need to use the colab i sent you the link
Either use that or our docs colabs guides
But the colab you sent before is outdated so it won't work
umbral acorn
Nice. Thank u
low shard
Nice. Thank u
Yw
ember lily
Seems fixed thank you
low shard
Ok good
wet cave
Does a longer dataset means longer epoch cycles?
rare gobletBOT
Ayo? @wet cave level 2 !!!
queen escarp
-colab
odd shale
Does a longer dataset means longer epoch cycles?
Not necessarily.
Everything will depend on the voice and tensorboard
brittle wing
Excuse me what wa the method to check lowest point of training how did I set smoothing
soft stratus
Click the download model? You used an outdated colab, thats why, you need to use...
It said that
low shard
It said that
You are using the new easygui colab right
soft stratus
You are using the new easygui colab right
Yes
low shard
Oh right your training
Ohh
Got it
Try to conver the dataset
To wav
As rvc has more compatibility with wav
And try to retrain like that
soft stratus
Try to conver the dataset
I tried but when i click download model said that
low shard
I tried but when i click download model said that
After you convert the dataset to wav, you would have to zip it again and restart training all over
soft stratus
After you convert the dataset to wav, you would have to zip it again and restart...
Do you mean process or extract
low shard
Do you mean process or extract
You need to zip your dataset and re run all the steps again
Ofc before convert it to wav
iron creek
my voice changer isn't working
supple inlet
Which music software was used to create the spectrogram image in this sample?
finite galleon
Which music software was used to create the spectrogram image in this sample?
spek
half cove
Which music software was used to create the spectrogram image in this sample?
oh fuck
wit
i took this image
finite galleon
i took this image
?
half cove
?
i remember screenshotting this for the website i think
leme check
lmfao
also its spek
hot lagoon
Hi there, I was just wondering something for back vocals because in the guide it says changing the hop length can someone explain why changing it helps the back vocals
half cove
yep
half cove
Hi there, I was just wondering something for back vocals because in the guide it...
where does it say this
supple inlet
i remember screenshotting this for the website i think
I borrowed this image from the following website.
https://docs.aihub.wtf/rvc/resources/datasets/
I apologize for using it without permission."
hot lagoon
where does it say this
half cove
I borrowed this image from the following website.
https://docs.aihub.wtf/rvc/res...
nah the website is public
i just remembered that i took it
hot lagoon
Yeah, but I don’t really understand the second option
hot lagoon
half cove
Yeah, but I don’t really understand the second option
mangio crepe is a option on mangio and applio to change hop size for crepe
hot lagoon
mangio crepe is a option on mangio and applio to change hop size for crepe
I understand the concept of both of them. I just don’t understand why changing or/converting the back vocals using the option is a option.
half cove
I understand the concept of both of them. I just don’t understand why changing o...
🤷♂️
just ignore it ig
hot lagoon
I guess you’re right I guess the best options are either recorded yourself and convert or using a voice synthesizer
strong fulcrum
Hello, how can I convert an audio file using RVC for mac through like command line or python or something like that? Is there like a pipeline or something?
iron creek
can someone help me with my vc changer
shy tiger
guys, can somebody please explain or send a guide, how to check model for overtraining?
cant find any info about it here
brittle wing
https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#tensorboard
I already found it but thanks
sinful nexus
-colab
scenic linden
how would you train joe bidens voice
trim nova
my output has this slight electronic sounding part on a certain part of the audio, when saying "S" on a higher pitched noise.. any possible fix for this?
wispy abyss
Any decent devs looking to join me in creating a completely new and revolutionary AI software. Hit me up I private and I’ll tell you more (there’s money that can be made 💰😉)!
brittle wing
Is it at the lowest point of overtraining
I exported it at 500 should I continue or stop training already
pastel oak
Is it at the lowest point of overtraining
Continue, OT is not clear yet
brittle wing
Continue, OT is not clear yet
Oh but but I'm not sure
I didn't set the correct smoothing
And the device's battery died
pastel oak
send new screenshot with 0,987
brittle wing
I'm continuing from another device on my phone so wait
split frigate
some1 told me if you dont have the stop a model train option, just force it close
i forced close after 800 epochs and there is no weights file,..
brittle wing
can someone help of Ilaria rvc?
pastel oak
some1 told me if you dont have the stop a model train option, just force it clos...
before you train rvc theres some settings to save a weight file every <NUMBER> epoch, did you not ave that enabled
brittle wing
split frigate
before you train rvc theres some settings to save a weight file every <NUMBER> e...
i didnt
its defualt at diasapled
brittle wing
https://docs.aihub.wtf/rvc/cloud/ilaria-rvc/
when i execute Start cell,
brittle wing
when i execute Start cell,
Traceback (most recent call last):
File "/content/Ilaria-RVC/app.py", line 1, in <module>
import subprocess, torch, os, traceback, sys, warnings, shutil, numpy as np
ModuleNotFoundError: No module named 'torch'
pastel oak
its gone then 💀
brittle wing
and no link appears
split frigate
its gone then 💀
i still cannot open tensor even if i downloaded Visual studio
pastel oak
idk
fickle tundra
hey uh can anyone tell me how to actually convert an audio file to 40000? i can only find 48000 or 44100
or 32000
brittle wing
Traceback (most recent call last):
File "/content/Ilaria-RVC/app.py", line 1, ...
can someone help this
pastel oak
on tensor board just see the highest eboshe and take it right?
basically u take this one the lowest
hard halo
Can't manage to remove the adlibs with BVE or DeNoise, tuned up the aggression settings as well and still get the same result.
What can I do?
fickle tundra
there is no sampling rate of 40000
well then why do they put 40000 there? is 40000 really 44100?
rare gobletBOT
Ayo? @fickle tundra level 7 !!!
split frigate
basically u take this one the lowest
the lowest 1 is the best 1 ?
pastel oak
the lowest 1 is the best 1 ?
ye
split frigate
i thought its the higher 1
brittle wing
well then why do they put 40000 there? is 40000 really 44100?
idk maybe 40000 means 44100
pastel oak
i thought its the higher 1
split frigate
no
https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#--monitoring
i still cannot open it, just wondering
pastel oak
idk
fickle tundra
hey i tried to upload custom pretrains and it came up with error (im using applio 3.0.9)
nvm i see the problem
my computer didnt let the thing move the files
so i did it manually
ok frick this im going to my old applio version atleast it worked
wet cave
Why does my rvc webui model inference output sound like this? What did I do wrong?
fickle tundra
Why does my rvc webui model inference output sound like this? What did I do wron...
what audio did you put in
wet cave
Just a guy talking. It sounds normal with the default models, such as kikiV1.
fickle tundra
Just a guy talking. It sounds normal with the default models, such as kikiV1.
oh ok i thought you left instruments in there or something
idk then
split frigate
what is best is the best between the 3 train programs?
or most use, idk how to say it
pastel oak
what is best is the best between the 3 train programs?
mainline
brittle wing
idk
Here's a detailed screen recording
Is it at it's lowest point or not
Cause I think it already is
From what I'm seeing here
brittle wing
lowest point at ~17k id say yea
There's a low point on the screen recording
brittle wing
yes i just answered that by saying its at 17k
Should I stop training?
pastel oak
yes
pastel oak
yes
brittle wing
Thanks for your assistance
brittle wing
you know where to find everything else ?
The files are in my drive
rare gobletBOT
Ayo? @brittle wing level 5 !!!
pastel oak
Will it sound robotic
Get the PTH file of your voice model which is before the lowest point. so you have to zoom into tensorboard, look for the number (lets say its 17.5k). then you select the model that is closest to that number
royal frost
i dont know if this is the right channel, i have 0 experience with RVC but i was wondering if someone would like to help out and run a 1 minute audio clip through the software using an AI voice model of rosario dawson. would someone like to help me out?
brittle wing
So basically I can send it to the person who requested that...
pastel oak
it should not sound robotic if your dataset is good
pastel oak
So basically I can send it to the person who requested that...
you should test it first probably but yea lol
brittle wing
Get the PTH file of your voice model which is *before* the lowest point. so you ...
I already exported at 500 epochs...
pastel oak
I already exported at 500 epochs...
noo you should have enabled the option to save weight files every <NUMBER> epochs. If you still have the D and G path, select the one before the lowest point, quickly run it so it saves the needed epoch
brittle wing
I just added 510 for the demonstration cause the other device is out of battery
pastel oak
i dont know if this is the right channel, i have 0 experience with RVC but i was...
yea sure send me everything you need
brittle wing
noo you should have enabled the option to save weight files every <NUMBER> epoch...
How
Save latest ckpt
brittle wing
I already enabled it
royal frost
yea sure send me everything you need
awsome! ive send you a DM!
brittle wing
Can you check which G and D files you have?
pastel oak
You unfortunately messed up, you needed checkpoints saved which you dont there :/
brittle wing
G_2333333
D_2333333
brittle wing
You unfortunately messed up, you needed checkpoints saved which you dont there :...
What
I had save latest checkpoint enabled
pastel oak
I had save latest checkpoint enabled
This has to be set to NO
If you set it to yes, it only saves the last
brittle wing
This has to be set to **NO**
I didn't use this colab
Bit the other one probably had this option too
What should I do now
pastel oak
start from the beginning 💀
you can try to see how the 500 epochs sounds and see if youre happy with it ig
fading notch
Where can I get Pokemon's Japanese rvc models ?💧
I can't found their💦
brittle wing
you can try to see how the 500 epochs sounds and see if youre happy with it ig
Is there a chance it won't sound robotic
Caus I spent so many days I have a personal life and other stuff to do
split frigate
you can try to see how the 500 epochs sounds and see if youre happy with it ig
in your opinion
which epoches perfect to put?
pastel oak
Is there a chance it won't sound robotic
i cant say yes or no to that, you just have to see how it sounds
brittle wing
in your opinion
which epoches perfect to put?
It's not a question of opinion.
pastel oak
in your opinion
which epoches perfect to put?
you just put 1000 epochs and train until tensorboard finishes. it could be 200 epochs, could be 500, always different
split frigate
It's not a question of opinion.
i usually use 500
pastel oak
make sure to save weights every 10 epochs or something
brittle wing
make sure to save weights every 10 epochs or something
That's what I did too
split frigate
700 fair enough i think
pastel oak
That's what I did too
wait so you have the weights then?
pastel oak
That's what I did too
its called something like
(as an example)
elena_260e_14000e.pth
brittle wing
wait so you have the weights then?
Where are they stored
pastel oak
can you check how many you got of those
brittle wing
can you check how many you got of those
In which folder is it on the file xplorer
Cause there's no such folder on my drive
brittle wing
i dont know how colab works
I don't have that on my drive
What do you use for training
RVC DISCONNECTED
But I closed the Colab
I'm testing the model
umbral acorn
make sure to save weights every 10 epochs or something
Hi
Is this overtraining?
foggy folio
guys will a 25 second dataset destroy me?
it is vulture from across the spider-verse
his lines are well. only at the start of the movie
azure marshBOT
RVC Guides (How to Make AI Cover)
thin stump
Hi
Is this overtraining?
"loss/g graph having high loss indicates that generator is uncapable (at the moment) of making realistic enough data. this usually occurs once the overfitting (overtraining) happens" Can you see what happened to d/total, kl, mel at 3k steps? You only need to type it in
thin stump
guys will a 25 second dataset destroy me?
I did a 35 second dataset and it came out clean but i'm thinking of switching to gpt sovits if I can figure it out
It's for tts only by the way. RVC is more for singing
foggy folio
I am using rvc
rare gobletBOT
Ayo? @foggy folio level 7 !!!
foggy folio
I don't think vulture has been made yet
thin stump
I don't think vulture has been made yet
Same as bane on the DC side. I think most youtubers use elevenlabs anyways like the one that features the spider-men
gusty kestrelBOT
@storm blaze, I have found 3 results that match your search!
Requested by tunc99
foggy folio
Same as bane on the DC side. I think most youtubers use elevenlabs anyways like ...
dc better moment
I'll make bane possibly
hot lagoon
Hi
Is this overtraining?
Well, before I answer that is ignore outliners on and is your smoothing set to 0.987
thin stump
Well, before I answer that is ignore outliners on and is your smoothing set to 0...
I thought that was common knowledge. Good call
hot lagoon
I thought that was common knowledge. Good call
Well, I’m guessing since they don’t have the model maker role that they just join the server or just now started making a model that’s why I ask
Sorry, my grammar is bad today
umbral acorn
Well, before I answer that is ignore outliners on and is your smoothing set to 0...
Yes
foggy folio
WHY SO MUCH PEOPLE USING MV SEP RIGHT NOW??
hot lagoon
Yes
I also want to ask how was the data set? How many batch size you put and what it looks like it is overtraining because it keeps going up which is bad for the model. I see the lowest point that you’re at is probably around 2K so you could try that one.
hot lagoon
WHY SO MUCH PEOPLE USING MV SEP RIGHT NOW??
No idea
split frigate
sry for repeat but this is annoying, how can i fix it
hot lagoon
Yes
The reason I keep asking is if your data is about two minutes then batch size 4 if it’s like more than two minutes than batch size 8
umbral acorn
I also want to ask how was the data set? How many batch size you put and what it...
My data is 17mins
Batch size is 8
So i should stop training?
Im using rvc disc
Do you know how to choose the lowest point of it ? 🥺
rare gobletBOT
Ayo? @umbral acorn level 3 !!!
foggy folio
The reason I keep asking is if your data is about two minutes then batch size 4 ...
if it is under 2 minutes?
hot lagoon
if it is under 2 minutes?
Someone told me that two minutes for bat size 4 it was a helper so
split frigate
sry for repeat but this is annoying, how can i fix it
its the tensorboard
hot lagoon
My data is 17mins
Batch size is 8
So i should stop training?
Im using rvc disc
...
Yes, if you do alt and scroll, then you could find the lowest point of the graph
Or find one that’s closest at least, but that depends on your save frequency
pastel oak
Maybe im not experienced enough just yet, but I think 2k steps for a 17 minute dataset is very rare? I'd go like at least 15k+ into overtraining to make sure my lowest point is actually the lowest or not?
thin stump
My data is 17mins
Batch size is 8
So i should stop training?
Im using rvc disc
...
Post your d/total, kl, and mel so that way we can see it
rare gobletBOT
Ayo? @thin stump level 19 !!!
cold vessel
Heyy does anyone know any free text to speech websites or apps?
hot lagoon
Maybe im not experienced enough just yet, but I think 2k steps for a 17 minute d...
I mean, I train about like a 30 minute data one before the lowest I could find that’s closest to the lowest was at 150 epoch
pastel oak
I mean, I train about like a 30 minute data one before the lowest I could find t...
Gotcha yea makes sense
pastel oak
If i could kiss u i would
hahahaha
mild glen
someone please help
thin stump
someone please help
That's not how it works. You need the added index so if a person didn't zip that in a model folder you can't use it
hot lagoon
someone please help
Correct me if I’m wrong but only make PH and index there’s no way to convert each other
pastel oak
Wait so how do i use the text to speech option?
i never used it myself, the guide shows a few TTS apps and online ways of using it
@split frigate do you use RVC disconnected?
umbral acorn
Post your d/total, kl, and mel so that way we can see it
You mean this one??
thin stump
type d/total in your tensorboard
pastel oak
Do you run rvc on your computer or your browser
thin stump
and you can share your whole screen if you have the snipping tool https://support.microsoft.com/en-us/windows/use-snipping-tool-to-capture-screenshots-00246869-1843-655f-f220-97299b865f6b
pastel oak
wait no that question is also bad
do you have rvc downloaded on your pc or do you use Colab
split frigate
Do you run rvc on your computer or your browser
if i press on go-web it move me to browser
pastel oak
if i press on go-web it move me to browser
ok so on pc
where did you put TensorVENV.bat ?
what folder
pastel oak
thats the mistake
put it in RVC folder
put it where go-web is in, same folder
and then start tensorvenv
split frigate
i putted it
pastel oak
now start
split frigate
it is
umbral acorn
type d/total in your tensorboard
This is d/total
pastel oak
your rvc is not unzipped?
thin stump
This is d/total
and type kl, that's all I need to see
split frigate
its unzipped
split frigate
i will redownload it
umbral acorn
and type kl, that's all I need to see
split frigate
idk what the issue is then sry
nvm its work
I just start it on the file without a shortcut and as an administrator
my bad tysm
thin stump
Has g/total reached a new low at 18k steps?
quasi flicker
C:\Windows\system32>MMVCServerSIO.exe -p 18888 --https false --content_vec_500 pretrain/checkpoint_best_legacy_500.pt --content_vec_500_onnx pretrain/content_vec_500.onnx --content_vec_500_onnx_on true --hubert_base pretrain/hubert_base.pt --hubert_base_jp pretrain/rinna_hubert_base_jp.pt --hubert_soft pretrain/hubert/hubert-soft-0d54a1f4.pt --nsf_hifigan pretrain/nsf_hifigan/model --crepe_onnx_full pretrain/crepe_onnx_full.onnx --crepe_onnx_tiny pretrain/crepe_onnx_tiny.onnx --rmvpe pretrain/rmvpe.pt --model_dir model_dir --samples samples.json
foggy folio
-dataset
azure marshBOT
-dataset
- Perfecting Audio Isolation on Low-End Rigs: A Practical Guide, by Litsa The Dancer and Faze Masta Google Docs
- Gathering and Isolating Audio, by SCRFilms Google Docs
- Creating Datasets for RVC using iZotope RX, by LUSBERT
Rentry - Vocal Mixing Tutorial, by Roomie YouTube
brittle wing
How do I bring the model back to 450 epochs
brittle wing
idk what the issue is then sry
Oh well the dataset the person gave me was already flawed
umbral acorn
Has g/total reached a new low at 18k steps?
I dont think so 🥲
This is so confusing…
brittle wing
I'm confused now
brittle wing
I dont think so 🥲
This is so confusing…
Did you go to the correct graph
umbral acorn
Did you go to the correct graph
This one is g/total
UHHHHHHH
Maybe first one i took is from grad_norm_d and i didnt realize!!
brittle wing
This is how you access it
foggy folio
-uvr
brittle wing
How do I go back to lower epochs
thin stump
Yup, 6k steps is where you would stop. It's overtraining throughout 8k-18k
brittle wing
Yup, 6k steps is where you would stop. It's overtraining throughout 8k-18k
Wait what
thin stump
I saw their kl and d/total. All of them are going down except g/total
brittle wing
RVC DISCONNECTED doesn't have the feature to save everything according to the save frequency to save at specific epochs like the RVC with persistent storage
So that's a problem I overtrained
umbral acorn
I saw their kl and d/total. All of them are going down except g/total
I was dumbbb 
sorryy
Also thanks for your helping
thin stump
I was dumbbb <:boohooh:1176674698629750975> sorryy
Also thanks for your helping
I made the same mistake too when I was looking at grad the whole time lol. Just make sure 6k is the lowest dip and not 2k-5k since it's cutoff from the picture
brittle wing
I made the same mistake too when I was looking at grad the whole time lol. Just ...
How do I revert to a previous amount of epochs
Back
Like I am at 500 I wanna go back to 470-450
thin stump
Like I am at 500 I wanna go back to 470-450
https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/#6-export-model I never did this before so i'm not the person to ask
and i'm pretty sure it gives you the files for 470-450 epochs unless i'm wrong
brittle wing
and i'm pretty sure it gives you the files for 470-450 epochs unless i'm wrong
It didn't but I'll check the guide
I have read this guide like 10 times there's no information there
Also can someone delete my old posts of models from my old account
split frigate
is RVC GUI important to get best quality for a model?
split frigate
I dont think so 🥲
This is so confusing…
mine shows like this.
is there something wrong ? or its normal, yours more details
thin stump
mine shows like this.
is there something wrong ? or its normal, yours more deta...
too early to say
split frigate
too early to say
ty
storm holly
is it normal that it goes off the chart
thin stump
is it normal that it goes off the chart
double click either one of them so it can zoom out
storm holly
here's how it looks like when i put it to scale
split frigate
now there are 2 charts what should i follow between them? to pick lowest epoch
storm holly
and turn off the eval checkbox
of course the latter does nothing
thin stump
now there are 2 charts what should i follow between them? to pick lowest epoch
https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/#-graph Both of you need to read this. It's going to train in the background so no worries
brittle wing
HOW DO I GO BACK TO 450 EPOCHS
The dataset already had some flaws like reverb and stuff
It wasn't made by me either I know HOW to make a decent datase
storm holly
i gave it a limit of 1000 epochs but i'm not sure it's gonna be enough
brittle wing
On which pitch extraction algorithm do I train someone's flawed dataset
rare gobletBOT
Ayo? @brittle wing level 6 !!!
brittle wing
Heyyy
foggy folio
how much epochs should I use for a 27 second dataset
fluid matrix
if its flawed, then it doesnt matter lol
fluid matrix
how much epochs should I use for a 27 second dataset
use ov2 pretrain and do 200 or something, check with tensorboard
thats not a lot tho, make sure its clean at least
foggy folio
I've really tried to clean it
fluid matrix
i gave it a limit of 1000 epochs but i'm not sure it's gonna be enough
1000 epochs is a lot, dont do that
more doesnt equal better, always check for overtraining with tensorboard
storm holly
more doesnt equal better, always check for overtraining with tensorboard
that's the thing
it's only gone down
line doesn't show signs of stopping
storm holly
can you show the graph?
storm holly
and how long is the dataset
3 minutes
storm holly
and why is only one of them enabled
it was @thin stump' idea
soft stratus
You need to zip your dataset and re run all the steps again
I tried but said error again
low shard
I tried but said error again
When it says error, show the output in the google colab cell
soft stratus
When it says error, show the output in the google colab cell
I saw but it said okay
low shard
I saw but it said okay
Wdym it said okay? Show a screenshot
Like
Click train
Then when it appears the error
Show the output in the colab
soft stratus
Show the output in the colab
And colab said nothing
brittle wing
if its flawed, then it doesnt matter lol
I don't make such type of datasets 💀
I mean the latest I made was clean to perfection this one has a reverb and it sounds weak at 400 epochs
brittle wing
A little reverb
fluid matrix
still
brittle wing
They're like 127 samples 💀
I don't have time to filter that out
brittle wing
So basically I trained it at 400 already the first time and it sounded weak.
fluid matrix
you can train it as long as you want, it wont improve the model if the dataset has reverb in it 😭
brittle wing
you can train it as long as you want, it wont improve the model if the dataset h...
It hit it's lowest point at 500
fluid matrix
I would cancel it and retrain after you cleaned it up tbh
denoise too if you havent
brittle wing
I would cancel it and retrain after you cleaned it up tbh
It's someone else's dataset 💀
I'll merge and unreverb
Lol
Im wasting my time it's ridiculous
I don't wanna do it anymore
brittle wing
What if use crepe or harvest on the flawed dataset
Cause it sounded weak with rvmpe
The singing sounded so unprofessional
So after merging and unreverbing, do I have to zip it?
lapis fox
Does anyone know why my audio file is empty when I save it? I can record 1minute and it just ends up being 1kb with no audio
fluid matrix
What if use crepe or harvest on the flawed dataset
please dont use harvest
and i keep saying it will sound bad if there is reverb
a bad dataset will sound bad no matter what
brittle wing
Nope I won't
brittle wing
always use rmvpe or crepe/mangio crepe
My first model was trained with crepe
It turned out perfectly
fluid matrix
well yeah crepe is still fine
it has its uses
but anything other than that is useless now
lapis fox
Does anyone have a good voice model that isn't an anime uwu girl, just like normal european or american english?
cant find one in #1175430844685484042 🤔
scenic linden
does anyone have a good voice model for joe biden
brittle wing
well yeah crepe is still fine
It has a reverb that cannot be removed 💀
Not a reverb that is from filtering out vocals more like one that is added with an audio program
median shuttle
c'e' qualcuno italiano che mi possa aiutare gentilmente?
brittle wing
c'e' qualcuno italiano che mi possa aiutare gentilmente?
Non sono italiana ma
brittle wing
ciao elena mi daresti una piccola mano
Si
median shuttle
guarda entri dento max 2 cosi ti mostro @brittle wing
brittle wing
I Will Just refrain It at 450 epochs and end it there
brittle wing
guarda entri dento max 2 cosi ti mostro <@456226577798135808>
Dove entro
loud bolt
can anyone help me how to voiceover smt? im new to the latest collab update and idk what to do anymore, it changed so much
fluid matrix
@median shuttle dont use rvc gui
you followed a old youtube tutorial probably
-rvc
fluid matrix
foggy folio
how do I know what amount of epochs to use?
brave garnetBOT
Made with love by Ilaria 💖
⠀
Google Colabs 
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
hoary zephyr
hey whats a good setup for local training, i have a 4 core cpu a 1070 and 32 gigs of ram. would that be good for training?
low shard
And colab said nothing
When you run the ui, under it theres like the ouput, so after you click train and it aays the error there should be the output of the error, maybe @quick trench could know a bit more (sorry for ping)
foggy folio
how do I pause and resume in applio?
(oh and also how do I make private models in weights.gg)
sly tundra
khz
brittle wing
oh my DAYS
why does this keep happening bruh
erm
the audio destination may be too long
it's 44 min
foggy folio
I'm making a black cat model from marvel's spider-man remastered
I know it already exists but eh
brittle wing
it's 44 min
this was it
plush compass
Is it better not to use harvest anymore? Is rmvpe unconditionally better?
brisk oak
can anyone direct me to non robotic sounding female voice?
plush compass
yes
Harvest best expresses the tone, with high notes and pronunciation being muffled.
I heard that mangio-crepe expresses the tone well, the high notes are not broken, and the pronunciation is accurate.
Is rmvpe a combination of these two advantages?
brisk oak
what does clean data mean
pastel oak
Harvest best expresses the tone, with high notes and pronunciation being muffled...
combination of pm and crepe but fast for both training and inference. its most faithful to the voice model
rare gobletBOT
Ayo? @pastel oak level 20 !!!
plush compass
combination of pm and crepe but fast for both training and inference. its most f...
Thanks for the advice
rare gobletBOT
Ayo? @plush compass level 10 !!!
pastel oak
definitely the best option
pastel oak
what does clean data mean
your dataset needs to be clean with no background noise etc. leo mentioned it in the message after
blissful dagger
guys what application do you guys use to generate the ai edited sounds
is applio good?
brisk oak
your dataset needs to be clean with no background noise etc. leo mentioned it in...
so just no background noise thats it? whats etcs? what else is there I could fix? except background noise?
pastel oak
so just no background noise thats it? whats etcs? what else is there I could fix...
noise, reverb, "popping sounds" (limit vocals to a specific volume for example to counteract that, i think)
just the usual that prevents a clean voice
plush compass
combination of pm and crepe but fast for both training and inference. its most f...
I'm sorry, but is it okay if I ask one more question?
I'm learning rapper AI.
Wouldn’t it be better to exclude auto-tune vocals from the dataset?
pastel oak
I'm sorry, but is it okay if I ask one more question?
I'm learning rapper AI.
...
I personally would exclude them to have more flexibility in post mastering and, if i have to, manually add auto tune in post prod if needed. Makes most sense to me, but I could be wrong
Plus mixing normal vocals and sudden autotuned-vocals might mess things up, as it might sound like 2 different people, and from the few practises that people have done, it's recommended to, if you really want to add multiple people to a dataset, do at least 3, as only 2 would cause issues. So in this case if the autotune vocals sound way different to the original, it'll come out as a mess
plush compass
I personally would exclude them to have more flexibility in post mastering and, ...
Wow... You organized it so kindly. Thank you so much.
pastel oak
Wow... You organized it so kindly. Thank you so much.
Take it with a grain of salt, I'm not an expert on the technical stuff, hope it was helpful
plush compass
thx
worthy tulip
@proven hill do you know how to fix the model error download status, i know the model isn't corrupt it works fine on other colabs
unreal lichen
hello i want uvr 5 colab
azure marshBOT
RVC Guides (How to Make AI Cover)
brave garnetBOT
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't inclued a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
sage dragon
sand jasper
Hello! I posted a forum about spectrogram, it would be really thankful if you stop by and leave any comments… Hope you have a nice day! Thank you!
soft stratus
When you run the ui, under it theres like the ouput, so after you click train an...
Here this problem
muted fiber
Hey guys how do I download Applio again? Sorry I had to memory wipe my computer recently
humble jetty
muted fiber
How do I get the same Applio that worked on a website tab
strong fulcrum
How do I get the same Applio that worked on a website tab
muted fiber
ty
brave garnetBOT
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't inclued a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
steel forge
This guide's batch files are offline https://docs.google.com/document/d/1ihvqWnSyabvZ9jX6BsOxEY1FpplwjY3Jse3she_VO6I/edit?usp=drivesdk
@proper shale
idk who's doc this is
cc @topaz oar
obsidian breach
yo i need help quick its for a yt youtube vid what is the best e girl rvc i need one to troll some people reply quick
muted fiber
when I run .bat or whatever it says it's on the Onedrive, and it needs to be moved somewhere else. What if Everything on my computer is Onedrive and nothing else?
soft stratus
I want anyone to create anyvoice model temporary
dire vapor
Can I use rin_e3 in other languages?
peak vault
Hey I'm pretty new to this so I have a question. I want to make a parody song of something, but I want to keep the original singer's voice. (aka keeping everything the same except the lyrics) Is there a way for me to do this?
full turret
-colab
thin stump
Hey I'm pretty new to this so I have a question. I want to make a parody song of...
like the channel There I ruined it? You can use rvc and convert a text to speech prompt but getting that tonally correct isn't simple so you have to be the one to sing it
peak vault
like the channel *There I ruined it*? You can use rvc and convert a text to spee...
Yeah kinda. 1. Where do I access this? 2. Does it auto-match the TTS with a tune, or would I have to do that myself?
thin stump
Yeah kinda. 1. Where do I access this? 2. Does it auto-match the TTS with a tune...
- You need RVC to create covers (Mainline, Ilariarvc, tryreplay.io) etc. https://docs.aihub.wtf/essentials/how-to-make-ai-cover/
- You already have a voice model downloaded either weights.gg and #voice-models (check the quality before using)
- There's no tts that can currently do that. The singing is done manually and they mixed it with the DAW of their
choice like FL studios etc. You can hear a sample of what https://www.fiverr.com/justvicco has done
peak vault
Alright, thank you so much
rare gobletBOT
Ayo? @peak vault level 1 !!!
thin stump
Alright, thank you so much
Yw, I think that pretty much covers it. I'm not sure if w-okada is worth mentioning but that's more of a real time voice changer. They have a channel about it here #🔍│help-w-okada
forgot to link the TTS pages https://docs.aihub.wtf/tts/tts-tools/
strong relic
Hey guys i have everything installed yet i cant hear myself in the voice or anything, the RVC thing is all working fine except the part where i cant hear the kikv1 voice or anything
trim nova
basically u take this one the lowest
is this RVC? how did you get that visualization
strong relic
@dawn nebula
strong relic
read this pls
doesnt list my issue
rare gobletBOT
Ayo? @strong relic level 1 !!!
strong relic
@brittle wing
strong relic
Hey guys i have everything installed yet i cant hear myself in the voice or anyt...
i hear it for like a second sometimes if i stop and restart conversion
thin stump
@brittle wing Rx is good for cleaning audio and sucks at separation
rare gobletBOT
Ayo? @brittle wing level 1 !!!
thin stump
You can easily see the noise in the middle and audacity is at the basic level
Yeah, best $5 purchase I made
⛵ 🏴☠️
never heard of it so I can't help with that
I use Mvsep since it's quicker. I have a lots of audio that I need to clean and there's multiple steps I have to go through to prepare it for training. I don't have a gpu for UVR locally https://mvsep.com/en
- BS Roformer V2 (mvsep exclusive)
- Izotope to dereverb, spectral denoise and so on https://rentry.org/RVC-dataset-RX-updated
- back to MVSEP to UVR de-echo if I still hear it
- UVR denoise
np, i've been using it for both ai models and covers
azure marshBOT
RVC Guides (How to Make AI Cover)
strong fulcrum
Why do I get an SSL certificate verification error when I try to download the prerequisites from https://huggingface.co/IAHispano/Applio/resolve/main/Resources in the RVC_CLI repo? "I ran python main.py prerequisites"
dusk socket
how would i go about making model
strong fulcrum
Could someone help me setup either Applio or RVC_CLI for mac? thanks
thin stump
any youtube to wav works. The free-mp3 site (that gave out flac files) was shut down a week ago
Or if you recorded with OBS, you can do mkv files to flac. Idk if it makes a difference whether I do wav or not
pastel oak
is this RVC? how did you get that visualization
additional application called tensorboard, its for rvc
raw wedge
Yeah, best $5 purchase I made
The RX software is $5?
raw wedge
- BS Roformer V2 (mvsep exclusive)
- Izotope to dereverb, spectral denoise and s...
What do you use to separate main and back vocal? I find usually back vocal still remains with the BVE model
trim nova
additional application called tensorboard, its for rvc
Thanks, got it working.. does lower loss in grad_norm_d and g means better voice cloning quality?
pastel oak
Thanks, got it working.. does lower loss in grad_norm_d and g means better voice...
not sure about those two graphs. usually people look at g/total, d/total, /mel and /kl
as long as theyre sinking, its good
theres some guides explaining what they all do
thin stump
The RX software is $5?
Check dms
trim nova
not sure about those two graphs. usually people look at g/total, d/total, /mel a...
I'm not a pro but I dont think this looks right.. 🥲
rare gobletBOT
Ayo? @trim nova level 2 !!!
pastel oak
I'm not a pro but I dont think this looks right.. 🥲
interesting
do you have a clean dataset with noises, reverb etc removed and how long is it? out of curiosity
trim nova
interesting
do you have a clean dataset with noises, reverb etc removed and how...
I tried to clean it as much as I can, it originally had worse echo from the microphone. It's slightly under 7 minutes
Im on epoch 324 rn
pastel oak
I tried to clean it as much as I can, it originally had worse echo from the micr...
you can keep training to see if anything changes, else you can try out the model where it had its lowest point in g/total at 3k steps and see how it sounds like
is it smoothened out to 0,987?
forgot to ask
trim nova
is it smoothened out to 0,987?
I'm not sure what's this
proud elbow
I'm not a pro but I dont think this looks right.. 🥲
try audio labeling method in Audacity like this, also batch size 8 is typically good to go to help reducing mode collapses
trim nova
on the left side of the graph is a slider
adjusted it to 0,987 looks like this now
trim nova
try audio labeling method in Audacity like this, also batch size 8 is typically ...
This is for getting a clean dataset yes?
pastel oak
adjusted it to 0,987 looks like this now
Looks fine to me, keep training
proud elbow
This is for getting a clean dataset yes?
yea good audio quality and denoising could help too
trim nova
yea good audio quality and denoising could help too
I'll check it out... I not quite sure but this actually helped me https://audioenhancer.ai/
proud elbow
I'll check it out... I not quite sure but this actually helped me https://audioe...
I'm not a fan of Adobe enhance and any similar tools, but if it causes some kind of artifacts and noise, you have to remove it again
pastel oak
Oh so it's fine yea?
yuup
trim nova
Any recommendation on post-processing tool? to get smoother output?
also since I only have 7 minute of source dataset, how many epoch should I approx aim for? or what's the typical "its good enough" parameters I should look for
low shard
also since I only have 7 minute of source dataset, how many epoch should I appro...
there isnt a specific epoch, look up https://docs.aihub.wtf/rvc/resources/epochs--tensorboard/
pastel oak
also since I only have 7 minute of source dataset, how many epoch should I appro...
when g/total hits the lowest point and it starts rising back up again
you use the model closest to before the lowest point
trim nova
Thanks @low shard @pastel oak
low shard
yw
queen escarp
-colab
proper shale
<@622856015444049937>
use the ones from docs.aihub.wtf
proper shale
harvest
crepe
rmvpe
rvmpe best for everything
unless you're training and have a suuuuper clean dataset, you don't need crepe
brittle wing
ahah
queen escarp
how to fix this?
brittle wing
how to fix this?
.
queen escarp
so i just use local version
im on mobile, i can't use local 😭😭
rare gobletBOT
Ayo? @queen escarp level 2 !!!
queen escarp
i don't have one
brittle wing
If >=3: apply median filtering to the harvested pitch results. The value represents the filter radius and can reduce breathiness.
brittle wing
rvmpe best for everything
unless you're training and have a suuuuper clean data...
how about this? set 3?
pastel oak
how to fix this?
brittle wing
Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy:
guys how do you set the value of these two?
feral saffron
Hey can someone tell me what colab to use for training?
low shard
how to fix this?
click runtime-> disconnect and delete runtime -> re do all the steps again from the start
low shard
Hey can someone tell me what colab to use for training?
you can either use rvcdisconnected, or applio, or easygui colabs
feral saffron
Do you happen to have a link to those?
low shard
you can find all of them on our docs https://docs.aihub.wtf
feral saffron
Thank you
low shard
Do you happen to have a link to those?
easygui dont really got a tut yet in our docs tho
the others do
feral saffron
Thank you man I appreciate it
spice blade
RVC started producing this error, can anyone help solve it?
Traceback (most recent call last):
File "C:\RVC_new\runtime\lib\site-packages\gradio\routes.py", line 321, in run_predict
output = await app.blocks.process_api(
File "C:\RVC_new\runtime\lib\site-packages\gradio\blocks.py", line 1007, in process_api
data = self.postprocess_data(fn_index, result["prediction"], state)
File "C:\RVC_new\runtime\lib\site-packages\gradio\blocks.py", line 953, in postprocess_data
prediction_value = block.postprocess(prediction_value)
File "C:\RVC_new\runtime\lib\site-packages\gradio\components.py", line 2076, in postprocess
processing_utils.audio_to_file(sample_rate, data, file.name)
File "C:\RVC_new\runtime\lib\site-packages\gradio\processing_utils.py", line 206, in audio_to_file
data = convert_to_16_bit_wav(data)
File "C:\RVC_new\runtime\lib\site-packages\gradio\processing_utils.py", line 219, in convert_to_16_bit_wav
if data.dtype in [np.float64, np.float32, np.float16]:
AttributeError: 'NoneType' object has no attribute 'dtype'
trim nova
when g/total hits the lowest point and it starts rising back up again
you use t...
Have you ever encountered this error in Tensorboard? TypeError: MessageToJson() got an unexpected keyword argument 'including_default_value_fields'
pastel oak
Have you ever encountered this error in Tensorboard? TypeError: MessageToJson() ...
no clue
proper shale
If >=3: apply median filtering to the harvested pitch results. The value represe...
no need to use this
proper shale
Protect voiceless consonants and breath sounds to prevent artifacts such as tear...
in case there's a lot of artifacting, decrease that
else, ignore it
jovial vapor
guys help me pls ai voice changer dont work. pipeline is not installiazed
rare gobletBOT
Ayo? @jovial vapor level 1 !!!
pastel oak
guys help me pls ai voice changer dont work. pipeline is not installiazed
delete folder pretrain and model_dir and run start_http again
polar fossil
Hello everyone, you can help with the problem. Everything is working, the sound is on, everything is set up correctly, but the RVC voices do not work. Help please.
jovial vapor
delete folder pretrain and model_dir and run start_http again
i cant delete model dir idk why
pastel oak
i cant delete model dir idk why
did you close voice changer and the cmd it opens?
check task manager too for voicechangerclient
jovial vapor
did you close voice changer and the cmd it opens?
check task manager too for vo...
yes
jovial vapor
did you close voice changer and the cmd it opens?
check task manager too for vo...
Is it necessary to delete model_dir?
pastel oak
Is it necessary to delete model_dir?
just restart pc and delete the folder then
its needed to reinstall the files
azure marshBOT
RVC Guides (How to Make AI Cover)
jovial vapor
just restart pc and delete the folder then
its needed to reinstall the files
it doesnt work
can you check on demonstration?
pls
jovial vapor
just restart pc and delete the folder then
its needed to reinstall the files
shad, only u can help me
rare gobletBOT
Ayo? @jovial vapor level 2 !!!
jovial vapor
please!!
sly stump
Hi guys, Im wondering if someone could help me with the W-okada? It won't open when I press the start_httpps even though the program ran just fine when I opened initially
brittle wing
Hey guys! (sorry my english is bad, I'm french) I'm searching a link for a google colab place, to make RVC or Applio with a tunnel, like Ngrok, for working whit GPu whitout stop. Do you know where I can do that? There is a bug on mine. Inference it's impossible now.
umbral acorn
is it me or iliria rvc is working slow as hell
brittle wing
is it me or iliria rvc is working slow as hell <:nails:1159569314848972891>
You work on a colab ?
rare gobletBOT
Ayo? @brittle wing level 1 !!!
umbral acorn
You work on a colab ?
Yes
brittle wing
For me on Illaria, it's okay now
umbral acorn
The result is loading super slow for me
low shard
Hey guys! (sorry my english is bad, I'm french) I'm searching a link for a googl...
you cant use google colab gpu without stop
it has a free daily limit
low shard
is it me or iliria rvc is working slow as hell <:nails:1159569314848972891>
be sure you are using the free daily gpu
umbral acorn
be sure you are using the free daily gpu
should i make new alt for that?
low shard
Hey guys! (sorry my english is bad, I'm french) I'm searching a link for a googl...
you can use the new easygui colab https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scrollTo=Sb5fzhzEXK8X or either just use any other colab in our docs https://docs.aihub.wtf which are different based on your needs
low shard
should i make new alt for that?
if you finished your free daily gpu, yes, either that or just wait for tmr to re use the free daily gpu, or either use cpu on ilaria rvc which is way slower
low shard
There is a tunnel on it?
you mean the ui? yes its easygui
low shard
https://colab.research.google.com/drive/1nrw5Y37-UKaWJYEBUPe6sayYeuXuEXHU?usp=sh...
idk where u got that colab but no it isnt from here idek if that is updated
brittle wing
idk where u got that colab but no it isnt from here idek if that is updated
Yes, it's a french colab, and it work with a Ngrok key, but now doen't work
brittle wing
idk where u got that colab but no it isnt from here idek if that is updated
Thanks for all informations dear
low shard
Yes, it's a french colab, and it work with a Ngrok key, but now doen't work
you need a ngrok key only for our realtime voice changer colab, but for easygui and others we use gradio or cloudfare tunnel/ui
brittle wing
you need a ngrok key only for our realtime voice changer colab, but for easygui ...
Okay! thanks!!
brittle wing
you need a ngrok key only for our realtime voice changer colab, but for easygui ...
Gradio and cloudfare can they work for a good while? Or are they cut very quickly by disconnecting the GPU?
low shard
Gradio and cloudfare can they work for a good while? Or are they cut very quickl...
wdym by disconnecting the gpu? all colabs stops working if you disconnect the gpu runtime or you finish it
you mean when you get disconnected for disallowed code ?
soft stratus
wdym by disconnecting the gpu? all colabs stops working if you disconnect the gp...
Can you train a model for me temporarily
soft stratus
you mean when you get disconnected for disallowed code ?
When RVC fix
umbral acorn
if you finished your free daily gpu, yes, either that or just wait for tmr to re...
Alright
low shard
Can you train a model for me temporarily
no sorry i dont have the time
request models in #1159289738314919936
brittle wing
you mean when you get disconnected for disallowed code ?
Thanks for all that already, I won't bother you any longer. I don't really know, it's just that when I tried Applio, the GPU crashed within a few minutes and I couldn't do many inference attempts. I had to reconnect it each time and restart the entire loading process.
maybe I can work longer on Illaria, I'll try.
low shard
Thanks for all that already, I won't bother you any longer. I don't really know,...
thats very weird, @peak tusk the disallowed code issue should be fixed right?
sleek crown
Hey. I'm trying to use "RVC Crepe v2 + persistent storage / autosave" colab but it only runs the tensorboard but not the gradio UI. Is this normal? What should I do?
low shard
Hey. I'm trying to use "RVC Crepe v2 + persistent storage / autosave" colab but ...
send google colab link
low shard
https://colab.research.google.com/drive/1TU-kkQWVf-PLO_hSa2QCMZS1XF5xVHqs?usp=sh...
that colab is very outdated
dont use it
use the colabs in our docs https://docs.aihub.wtf or easygui
sleek crown
Thanks! I'll give it a try
rare gobletBOT
Ayo? @sleek crown level 1 !!!
low shard
yw
queen escarp
-colab
last patio
please someone tell me how i make a rvc model
tranquil trail
@red kayak
yo bro
is there a declipping tool
for getting vocals bc i noticed the kick or 808 kinda clips the voice
and it makes it sound staticy
any way to like remove that
denoise aint rlly help
red kayak
normalize
red kayak
in audactiy
yes
tranquil trail
yes
it
dont
help tho
it seems like it lowkey muffles it
bc of sometimes how hard the kick hits
are there any new settings for audacity
or are they all the same?
like in the old guide
strong fulcrum
Hello, could someone help me with setting up rvc cli? I’m getting some errors during setup
rare gobletBOT
Ayo? @strong fulcrum level 3 !!!
cyan mesa
Can you send a link to create an audio model and make an AI Cover?
azure marshBOT
RVC Guides (How to Make AI Cover)
rare gobletBOT
Ayo? @brittle wing level 2 !!!
odd shale
You must know how to use it.
It's only good if you know how to use it.
Don't put it on 9 or 10 of strenght on RX unless your audio has a fckton of reverb (if it's an dialogue audio)
If you're trying to dereverb an acapella, just use De-Echo-DeReverb model on UVR
wheat maple
hey anyone knows how to make a model?
dusk plaza
hey guys i keep getting errors when i try to convert my audio file to a voice model i choose
and when i download the voice model it says this
ERROR - The download failed. Check if the link is valid.
but the link is valid
and then the model shows up
and when i try to convert my audio
it says error
low shard
ERROR - The download failed. Check if the link is valid.
Send model download link
I dont know much about what is gaudio, but uvr lets you remove vocals, instrumentals, reverb, harmony and this all for free and also backing vocals
dull orchid
Resample the output audio in post-processing to the final sample rate. Set to 0 for no resampling: ( Can someone explain what this does exactly) ?
brittle wing
Is there any way to train models on mac m1?? locally
low shard
Is there any way to train models on mac m1?? locally
Nope
Just use google colab
But not locally
olive acorn
is there a discord server where i could ask questions about the voice changer itself? because i am encountering some issues at the moment
spare shale
olive acorn
ty
rare gobletBOT
Ayo? @olive acorn level 1 !!!
austere portal
hey my RVC launchers instantly open and close CMD after running them and doing nothing
upper furnace
because I do all the steps well and it gives me this and it doesn't start what do I do? (GPT SOVITS)
rotund needle
hey my RVC launchers instantly open and close CMD after running them and doing n...
how do you get RVC?
austere portal
how do you get RVC?
installed it manually again to update according to this: https://www.youtube.com/watch?v=YMxXtaiVHks
rare gobletBOT
Ayo? @austere portal level 1 !!!
rotund needle
installed it manually again to update according to this: https://www.youtube.com...
ok ty
upper furnace
?
outer stone
What is rmvpe vs. rmvpe_gpu? Is this RVC 2?
austere portal
ok ty
ah shit just realized wrong video, this is the one: https://www.youtube.com/watch?v=nXpBlC6OBw4&t=200s
any clue?
muted fiber
When i run the .bat file, it says my Applio is on the OneDrive Desktop, but my whole computer is on the OneDrive Desktop. How do I fix this?
👀
brittle wing
Do you know why Applio takes so long to install?
frank shuttle
-rvc
frank shuttle
-colab
static matrix
once i have a model how do i use it for making a song
indigo sand
Normally using one of the links above, but I'm personnally struggling to get any of them working atm.
glad flume
my step 2 "download the model" is running an error, does anyone know how to fix it?
indigo sand
i get the download to work but not the file conversion.
Not sure why everything is broken for me
glad flume
for those with rvc gui not working, try this one instead: https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb , i saw this tutorial: https://www.youtube.com/watch?v=73Tl_7GBZsA
azure marshBOT
RVC Guides (How to Make AI Cover)
karmic heron
Bruh whats wrong?
skill issue
i mean
get colab pro
so you dont get this message
errant stag
get colab pro
isn't t4 free?
wispy burrow
Hello everyone!
I am back! (had computer issues).
Now...what's new in the world of RVC (& UVR)?
Anything I missed?
Is it still Applio, Mangio & Regular RVC?
Did Applio's installer get "fixed".
What is this Ov2 & Ov2 Super I see on some Voice Models here?
Why are there drums in #1175430844685484042 ?
karmic heron
so it's just not free anymore
no i mean
colab is free
t4 is free
but if you pay for colab pro you can use webui
without these messages
errant stag
without these messages
so its just not free 💀
finite galleon
Hello everyone!
I am back! (had computer issues).
Now...what's new in the world...
Nothing new on rvc since its dead and yes its still the trios which all of them will stop updating soon. Ov2 and ov2 super are the same thing. Idk why there are drum models on #1175430844685484042 but its just drums trained on rvc.
karmic heron
so its just not free 💀
IT IS FREE.
WITHOUT webui is free
but WITH webui
you need to pay
if you break their tos they will ban you (using webui in google colab free tier)
wispy burrow
Nothing new on rvc since its dead and yes its still the trios which all of them ...
It's dying? Or are they moving on to something new/better?
Still trying to find out what this Ov2 is...
finite galleon
It's dying? Or are they moving on to something new/better?
Still trying to find ...
Rvc has no major updates anymore. They are moving to gpt sovits but its not voice conversion but tts. Ov2 is a training pretrain, i wouldnt use it if your dataset length is above 5 minutes
broken prairie
Hello everyone!
I am back! (had computer issues).
Now...what's new in the world...
It all started back in the days of the old AI Hub
I make RVC drum models because deepfake laws only protect voices
half cove
Of you mean colab vs local if its possible for you to use local then it's recommended
cinder canopy
thanks
soft stratus
Is RVC easy gui fixed
earnest sage
hi everybody, I was trying to create a voice model but I could not create. Can someone help me to create or create for me. I need that 🙂
I sent request but nobody tryied to help me
strong fulcrum
Hello, could somebody help me in the setup for either rvc_cli or applio on mac? Im getting errors and cant find solution online
pastel oak
hi everybody, I was trying to create a voice model but I could not create. Can s...
itd be helpful if you explained what doesnt work
earnest sage
itd be helpful if you explained what doesnt work
The website become disconneted again and again
I gave up to do that I think because this is so difficult for me I cant understand
I am not good at using pc enough
pastel oak
which one are you using send link
earnest sage
I need a model, I did setup the voice changer but I cant create voice model
sir if you are available, can you create for me voice model. I can send link that I am trying to create model(youtuber)
pastel oak
https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scroll...
i dont recognize this
please use one of the online versions of the guide
earnest sage
i dont recognize this
Alright. thank you so much
strong fulcrum
RVC disconnected or ilaria is good
Hello, do u know anything about rvc_cli or applio?
pastel oak
Hello, do u know anything about rvc_cli or applio?
I dont
pastel oak
Ok. What do u use for running rvc models?
i do it locally on Mainline
for colab i recommend one of the two i mentioned since theyre easy but i never used any colab so could be wtong
strong fulcrum
I see. I want to run rvc model locally, like applying rvc some audio file
strong fulcrum
i do it locally on Mainline
What is mainline?
pastel oak
What is mainline?
on the guide its called "mainline", i forgot the real name they just call it that
strong fulcrum
on the guide its called "mainline", i forgot the real name they just call it tha...
U mean on docs.aihub.wtf?
pastel oak
U mean on docs.aihub.wtf?
yed
finite galleon
on the guide its called "mainline", i forgot the real name they just call it tha...
mainline is rvc
rvc webui to be specific
strong fulcrum
rvc webui to be specific
ah i see. tho can i choose to run it without the webui? like from command or in python?
strong fulcrum
rvc webui to be specific
@pastel oak Also, do you know if this works for mac? Because the download linked from the guide to the huggingface was to the nvidia 7z. does the other versions work for mac like the beta or smth?
desert niche
-colab
teal dome
is something wrong with ilaria
(the tempfix colab)
ik you have to do the process twice since it fails the first time but now when i download a model it immediately says it was successful (suspicious, it should take a bit to download it) but there’s nothing in the list after refreshing
rare gobletBOT
Ayo? @teal dome level 2 !!!
proven hill
is something wrong with ilaria
no thanks im fine but thank you for your concern
proven hill
wait you’re a peach stan too omg what a slay
only bc of this
youre my new bestie
ill be your personal helper
what an honor, i know, a round of applause
now, are you using the colab?
raw violet
im having trouble on the Easy GUI site posting the URL of the models. When i try to run it, it gives me a error everytime on every model.
any assists?
teal dome
yeah, trying a 3rd time (Ilaria_RVC_Tempfix) and it’s at 5m so far
proven hill
im having trouble on the Easy GUI site posting the URL of the models. When i try...
colab link?
raw violet
colab link?
yes
proven hill
yeah, trying a 3rd time (Ilaria_RVC_Tempfix) and it’s at 5m so far
the hell is Ilaria_RVC_Tempfix
teal dome
idk i got it from here like a week or two ago 😭
raw violet
rare gobletBOT
Ayo? @raw violet level 1 !!!
indigo aspen
how do i continue training on rvc disconnected?
rare gobletBOT
Ayo? @indigo aspen level 2 !!!
proven hill
https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sh...
thats a super mega old link
raw violet
oooh
proven hill
how do i continue training on rvc disconnected?
did you save on google drive?
indigo aspen
yes
proven hill
should be the buttons to resume
raw violet
thats a super mega old link
can i have the new one ?
proven hill
can i have the new one ?
the one i sent above, coded by me
indigo aspen
should be the buttons to resume
do i have to re run this
proven hill
do i have to re run this
why not a simple screenshot
im sever dyslexic i cant see/hear this
teal dome
use this: https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo...
works like a charm thank u miss ma’am
proven hill
miss maam😭
finite galleon
ah i see. tho can i choose to run it without the webui? like from command or in ...
Yeah you can run it with python using the VC class and vc_single method from infer/modules/vc/modules.py
Not sure about commands tho
proud elbow
do i have to re run this
the training variables and then this
strong fulcrum
Yeah you can run it with python using the `VC` class and `vc_single` method from...
So even tho it says Nvidia when i download and guide says to run bat, i can ignore that and just run it thru python on my mac?
finite galleon
So even tho it says Nvidia when i download and guide says to run bat, i can igno...
Not sure on mac
I dont do rvc locally
strong fulcrum
I dont do rvc locally
what do u do? colab?
finite galleon
what do u do? colab?
Yup
proud elbow
what do u do? colab?
I do infer locally in cpu mode, colab for training only
strong fulcrum
Yeah you can run it with python using the `VC` class and `vc_single` method from...
and could u show what exactly u mean? i see infer_pack/modules.py
strong fulcrum
I do infer locally in cpu mode, colab for training only
i see. what do u do exactly?
finite galleon
and could u show what exactly u mean? i see infer_pack/modules.py
Yeah its the modules.py file
I forgot the exact path
strong fulcrum
Yeah its the modules.py file
is this the only file i need?
finite galleon
is this the only file i need?
The file will use other files to run it
Like pipeline.py
but modules.py just gathers everything up so you can just do vc_single and it will do inference
strong fulcrum
Like pipeline.py
i dont see pipeline.py
ah i think i do now. vc_infer_pipeline.py
finite galleon
Yeah that
strong fulcrum
Yeah that
tho i'm looking thru the modules.py file and it doesnt mention the pipeline file
finite galleon
tho i'm looking thru the modules.py file and it doesnt mention the pipeline file
There should be like a from vc_infer_pipeline code on modules.py
strong fulcrum
There should be like a `from vc_infer_pipeline` code on modules.py
nothin
finite galleon
https://huggingface.co/lj1995/VoiceConversionWebUI/blob/main/infer_pack/modules....
I was looking at the github page lmao
Theres a infer.py file there
Just use that
teal dome
would anyone know if there's a way i can re-organise this guy's model to work again https://www.weights.gg/models/clm72r77j0lfhcctcifydd9eh
it used to work with rvc even despite giving an error but as of now it actually doesn't work perioud
strong fulcrum
I was looking at the github page lmao
oh ic lol. which github page?
rare gobletBOT
Ayo? @strong fulcrum level 4 !!!
teal dome
i'm assuming it's because of how it's been organised here
finite galleon
Why is the code on the github repo and huggingface different
finite galleon
oh ic lol. which github page?
The rvc repo
strong fulcrum
Why is the code on the github repo and huggingface different
lmao idk
finite galleon
Ye
strong fulcrum
Ye
yea the structure is different. maybe came out different times?
huggingface one last updated about 4 months ago added file
finite galleon
yea the structure is different. maybe came out different times?
Possibly
strong fulcrum
Possibly
so if i were to use this repo as a whole would the environment work with mac?
like i mean not just take a couple files out like modules.py and pipeline
finite galleon
so if i were to use this repo as a whole would the environment work with mac?
Again i am not sure i dont use rvc locally
strong fulcrum
Again i am not sure i dont use rvc locally
okay alright
hm im trying to see if it works thru cli but the cli commands aren't workin
azure marshBOT
RVC Guides (How to Make AI Cover)
wispy burrow
would anyone know if there's a way i can re-organise this guy's model to work ag...
May depend a little on what RVC you you're using.
For Applio (local) the added_IVF*.INDEX goes in
Logs\ModelName
The ModelName.pth goes in
Logs\Weights
Those are the important ones.
But the rest too can go in
Logs\ModelName\
teal dome
i'm using the ilaria colab
forgor to ping oops @wispy burrow
so far i tried like 9 different ways of re-organisation and they all failed
wispy burrow
how much space does applio take?
I have one that almost 9GB without any models and such.
wispy burrow
i'm using the ilaria colab
Not familiar with Colabs.
Tried looking for a guide here but the link is broken...
teal dome
would anyone know if there's a way i can re-organise this guy's model to work ag...
would u @proven hill... happen to know what could be wrong here 😭
proven hill
would u <@740255187234717838>... happen to know what could be wrong here 😭
yes
inside the zip file there has to be ONLY a pth file and the added index
no sub folders
teal dome
will monty still work the same if i remove all that extra stuff?
moment of truth... lemme see how it goes
oh my god FINALLY
thank u times 2
rare gobletBOT
Ayo? @teal dome level 3 !!!
teal dome
i do wonder... what all those extra files were for if it works fine without them
wispy burrow
i do wonder... what all those extra files were for if it works fine without them
Those were created during the training of the model. (not even everything is there)
Those are not needed for Inferrence.
proven hill
thank u times 2
yay
if you need anythinf let me know
send me a friend request too
works only for you bc youre so real for peach
trim nova
@pastel oak hi, which value do I need to look at specifically the smoothed one or "value"
faint bloom
what am i doing wrong
i put in the audio and pressed the puttons labled 1 2 3 4 5
5 is always error
hollow tangle
can someone tell me what a index file is/
faint bloom
can someone tell me what a index file is/
no
rugged stirrup
Its normal just go to the "Logs" folder and download the index and then got to "Assets" folder and download the pth file
faint bloom
Its normal just go to the "Logs" folder and download the index and then got to "...
what
rugged stirrup
can someone tell me what a index file is/
It helps with the accent of the voice
faint bloom
Its normal just go to the "Logs" folder and download the index and then got to "...
i dont have no foldiers
rugged stirrup
i dont have no foldiers
You are using colab right?
faint bloom
ye
rugged stirrup
Go to the colab page and click on the little folder button if the folders arent open
faint bloom
