✨│ai-help

[analog obsidian]

he can't use mainline preprocessed slices because applio uses their own weird slice structure 😭

[glacial pollen]

You preprocess, then extract on mainline
and move that

from exp's dir back to applio

[glacial pollen]

🤔

oh

the naming?

[analog obsidian]

yea

it does not affect?

[glacial pollen]

I mean

you can just rename the folders

👀

[analog obsidian]

lmaaao sorry im restarted

[crude flame]

applio's rmvpe does use rmvpe gpu

[analog obsidian]

[glacial pollen]

0_gt_wavs - > sliced_audios
1_16k_wavs -> sliced_audios_16k
2a_f0 -> f0
2b-f0nsf -> f0_voiced
3_feature768 -> v2_extracted

should be it, ye

[coral frigate]

i tried to feature extract 1 50 min file and still got the same error. and this is after i just redownloaded applio

[analog obsidian]

wow now is a cudnn error

[coral frigate]

i move these to to my models folder in applio?

pretty sure that error was there before

[analog obsidian]

so weird so u got cudnn error and oom error

[coral frigate]

its so frustrating

[glacial pollen]

hmmm..

try to test 1 thing

and pick 3.7

you run it in cmd

py slice.py

tho

before you do

first make a copy of what you used in

[glacial pollen]

^

and the path in cmd should lead to that one

That'll give you uniform 3.7 sec segments

• discard the last one ( as in 99.99% of cases, that one isn't the length we want )

And then, attempt the extraction on sliced ones ( instead of 1 fused 50 min

[coral frigate]

so py slice.py then enter the ^?

[glacial pollen]

you can put the file to where you run the script in ( file you wanna slice )

'll output the folder with slices in that exact spot

Then, move the folder with slices to wherever you keep ur set and yea.. work on that one

[knotty moth]

do you still get oom error? then other possibilities: corrupt driver, BIOS, or ram/vram itself

[analog obsidian]

nope now he got cudnn error

[knotty moth]

have he tried the original mainline rvc?

[analog obsidian]

hes now trying what codename told him

slicing the file outside applio/mainline

[coral frigate]

when i run the python file. should i be seeing a gui or just open it in notepad and save the changes i make?

[glacial pollen]

oh

nono

[coral frigate]

srry

[glacial pollen]

you run a cmd window like so:

[coral frigate]

im trying to keep up

[glacial pollen]

that way

in whichever dir you are at given moment, if you type in cmd in the address bar, it'll open it up with the path set

in there you just simply type:

py slice.py

in essence, you run a .py ( python ) script using py ( python )
<executor> <script>.py

is the template ^ for easy memorization

[coral frigate]

do i add the audio file in the same folder as the python file?

[knotty moth]

btw is ai-testing chan gone or only for your role? I thought it should be kept public since the rvc project is open source

[glacial pollen]

mh?

I don't know nor see it the way normal users do so, even if something happened to the visiblity or access, I wouldn't know

[coral frigate]

Sorry for the delay. I had to install python and all the dependencies to run the code. Whenever I try to add the path to the wav file it gives me a permission denied error. I tried adding the path to the folder and also adding the path directly to the file

[glacial pollen]

uh

and script + file-to-slice in the same folder?

[coral frigate]

Tried that too. When I copy the file path it says file not found even though I’ve checked the file and the path I copy is correct

[glacial pollen]

is it wave tho

[coral frigate]

It is

I’m going to try splicing with audacity and see if that works

[glacial pollen]

well, in any case, that's weird as it shouldn't be doing that unless something's wrong with perms or python maybe? on your side

[coral frigate]

I don’t know. That’s a problem for another day. Today is just not my day

[knotty moth]

move the script and input file out of onedrive or program files/system directory

[low shard]

Yw

[twilit phoenix]

Hey guys, can anyone help me with rvc? Im using m2 silicone and whenever i tried using rvc it bugs out or just takes a really long time for my voice to be transferred and turned into the voice i wanted, and whenever i tried changing into a differrnt model it doesn’t let me (mind you im using the latest rvc demo for mac silicone)

[azure marsh]

Hey, Meruemu! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[low shard]

Are you talking about RVC, Applio or Wokada? You can send the guide/program link you're using

I also remind u Mac is not the best for AI, but u shouldn't have the switch model issue

[twilit phoenix]

Honestly idk which one, since im not good at computer stuff, but it says here i downloaded vcclient_mac_2.0.73-beta.zip

[twilit phoenix]

I dont really have anything else other than this mac, should I change and download other voice changer client

Or is it just gonna be that bad and laggy forever, or did I download the wrong rvc/applio/wokada for my m2 mac?

[low shard]

That's called Wokada as seen in the GitHub repo, its a program that uses RVC (Retrieval-based-Voice-Conversion, Speech To Speech AI type) models in real-time for calls

Don't follow yt tuts btw

[low shard]

Are you looking for realtime voice changer for calls or RVC (RVC is used for training, making models, and inference, using them, on pre-recorded audios)?

[twilit phoenix]

Ohh, okay I understand

[twilit phoenix]

For calls, discord specifically

[low shard]

Then yeah Wokada, I suggest you to use the Wokada deiteris fork (fork = modified version), the fork of deiteris has way better performance

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[low shard]

1st guide

[twilit phoenix]

Tysm! If there is another problem may I follow up with you? Thank you for giving me your time

[low shard]

Oh just seen this as I didn't get pinged lol,
Tbh I'm not a specific Wokada helper also bc I don't have a Mac, but u could ask in #🔍│help-w-okada

Also I'm glad I was helpful

[flint solar]

Can I use okada on the cloud

Nerve acc tried it

[low shard]

Yes

Did you check if your PC is good enough first tho ?

Local is better than cloud usually as cloud has GPU time limit

[flint solar]

It is but I’m on laptop rn

I don’t think it can run it

[flint solar]

It’s available js on colab?

Kaggle better

[low shard]

Maybe, the Wokada Deiteris fork is a much more performant version

You got 2 options:

• Run it locally (on ur pc) using the CPU mode of the wokada fork which has better performance https://rentry.co/ForkVoiceChangerGuide (but im not sure how good it's gonna be for the delay)
• Use cloud (remote good pc):

About Cloud, there are different services:

• Google Colabs (4 hours daily of free T4 gpu, easy to use, require only a google account) :
  • How to use Original W-Okada's Voice Changer Google Colab (has a Guide)
  • Modified W-Okada's Voice Changer Google Colab
• Kaggles (30 hours weekly of better GPUs, T4x2 & P100, harder to use, requires an account and a phone number)
  • Original W-Okada's Voice Changer Kaggle
  • Modified W-Okada's Voice Changer Kaggle

Not sure how good is the CPU mode, I think I heard @pastel oak saying its good for cpu but can't know

[pastel oak]

A laptop with no gpu can only have a shit cpu aswell, dont waste your time

I recommend fork kaggle

[flint solar]

It got a intel i7 8th gen

[coral frigate]

hi i was able to finally run the model by using mangio. but the slope on the tensorboard seems off for it only being at 25 epochs. should i just leave it to keep training?

[simple ore]

65k steps is just 25 epochs?

Assuming you did batch 8

that's about 17 hours of audio sliced with 3s chunks?

and you're doing a regular training?

[coral frigate]

yes thats about right

not sure about the chunks but the rest is correct

[coral frigate]

btw i really did do everything to get applio to work with my dataset but nothing would work so i had to bust out this relic again

[simple ore]

17 hours is way too much for regular training for a single voice using a pretrain

yeah, you can probably take a model saved around 40k steps and it should be fantastic

[coral frigate]

yesterday i was told that 16 is not enough without a pretrain

[simple ore]

if you train from scratch, no pretrain, then you need more

and you need more variety, not a single speaker

[coral frigate]

even if its only 25 epochs?

[simple ore]

with pretrain? yes

without pretrain it may be okay

[coral frigate]

i used the default for the pretrain that was loaded when you first open the app. guessing that means i did use a pretrain

[flint solar]

whats the math behind this

[simple ore]

as for applio not being able to handle 17 hours of audio.. it depends on whether you've selected a batch size and gpu caching that ran out of vram

[coral frigate]

no we went through a whole thing with like 3 others last night in the chat trying to troubleshoot it and nothing worked

[simple ore]

just a guess for batch size... but 1 step = batch size x average chunk of audio (3s)

[quasi dagger]

Is a dataset with 17h really worth it?

[flint solar]

probably nor

😭

[simple ore]

for training from scratch without pretrain, not really enough, for training with a pretrain it is a big overkill

[quasi dagger]

Longer I've seen was like 6-8h

[simple ore]

unless those 17 hours are 17 separate speakers

[quasi dagger]

And the result sounded bad in the end

[analog obsidian]

because it was trained without a pretrain

[quasi dagger]

Hey, Lyery!

[latent kettle]

hello @simple ore @nocturne mural whenever i try to uplaod .index (models from weights.gg) in applio i got an error " Applio_RVC_Fork\3.2.7\tabs\download\download.py", line 42, in save_drop_model
file_name.split("nprobe_1")[1].split("_v1")[0]
IndexError: list index out of range "
@low shard suggested me to put " added_IVF738_Flat_nprobe_1_NameOfModel_v2.index " in index and now its working.

[analog obsidian]

[low shard]

btw, he means an index from weights.gg

[analog obsidian]

seoul was trying to see if 9 hours was enough to train without a pretrain
spoiler: it wasnt

[flint solar]

shit takes weeks

[analog obsidian]

yea because you're supposed to use over 50 hours to create a pretrain

original pretrain took 4 days

[latent kettle]

i corrected it thanq for reminder

[coral frigate]

i pulled it and it sounds robotic. dunno if i should let it train more or idk

[flint solar]

and its the best pretrain yet

[simple ore]

i would not call it the best simply due to the noisy data

[analog obsidian]

not the best but the one that works better

[simple ore]

og pretrain has 109 speaker variety that helps a lot

[knotty moth]

nah you could never train a usable single speaker model without pretrain
||not even what that guy with frieren pfp had done||

[analog obsidian]

exactly, its not possible but he wanted to try it anyways lol
but in his defense, the model was made in a era where we did not knew about rvc in depth

[latent kettle]

@simple ore can you please fix it ??

[simple ore]

i dont know what download thing is doing

upload index to where?

[latent kettle]

latest version of applio

[simple ore]

what are you trying to do?

[latent kettle]

applio 3.2.7

[analog obsidian]

which batch size? 16?

[coral frigate]

8

[analog obsidian]

yep this explains

[simple ore]

if you want to use model from weights.gg in applio, all you need to do is download the .zip of the model and unzip it into Applio's logs folder

[latent kettle]

im trying to uplaod .pth and .index in applio 3.2.7 and it shows error in CMD

[quasi dagger]

I once was getting an error from a model from weights as well so I didn't use the index and it worked

[analog obsidian]

why are you using mangio over mainline anyways?

[latent kettle]

i just use download tab in applio.

[analog obsidian]

mangio is outdated and stopped getting updates since 1 year and a half

is not even up to latest mainline update

[coral frigate]

i only know of the two. idk what mainline is

[analog obsidian]

original rvc

unedited, no fork

[low shard]

i thought it was fixed for when u upload indexes from weights.gg

[analog obsidian]

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases/tag/2.2.231006

applio and mangio are edits of this

[coral frigate]

ok thanks. so 16 batch size with mainline?

[analog obsidian]

yes, your dataset is huge so i don't believe batch 16 is going to cause harm

[coral frigate]

thought lower batch size meant better quality. didnt think it would make the sound worse

[analog obsidian]

8 technically is better but it would take u 1 day or 2 for good results

is not

batch size is not a quality setting

bs 8 and 16 will sound exactly the same in your case

difference is that bs 8 will generalize better

[quasi dagger]

I didn't even know it had this drop feature, I just extract the zip (manually downloaded from weights) and put in the folder

[analog obsidian]

quality wise they are going to sound exactly the same

[coral frigate]

my tensorboard went flat after 25 epochs. but if youre saying that it shouldve taken 1-2 days makes me think i prob messed something up

ill retry with the new settings on mainline and see how tyhings go

[analog obsidian]

on big datasets this is normal

getting smooth graphs

actually on bs 16 you're going to get an even smoother graph

[simple ore]

also dont fucking use FP16 for training

that'll ruin everything

[analog obsidian]

fp32 will use more vram but u have a 4090 so it doesnt affect you

training is going to be more slower than the fp16 run

[coral frigate]

i didnt see anything about fp32 and fp16

[analog obsidian]

on mainline you can change this in config.py

[simple ore]

FP16 is how your model goes off the rails right into garbage territory

[analog obsidian]

self.is_half = false

[coral frigate]

is fp16 on by default? should i change the settings after i download it?

[quasi dagger]

Never saw that as well. Seems like it's in the config as Lyery said

[analog obsidian]

yes you have to disable it

is enabled by default

[coral frigate]

ok just switch it to false

[analog obsidian]

yep

[analog obsidian]

set to false as well

[coral frigate]

so mainline, 16 batch size, disable the setting in the config and keep the training running even if the tensorboard has gone flat. sound right?

[analog obsidian]

pic got cropped lol
yea this is self explanatony

[analog obsidian]

yes

flattening in big datasets is normal

i would say you will get usuable results at around 24 hours or so of training

(because the dataset is big)

[latent kettle]

Ohhh.. I'm just informing the developer that this feature is broken so he can fix it for others..

[quasi dagger]

Sounds good. @simple ore can you address this?

[analog obsidian]

but occasionaly check the graph, if g/total is fluctuating a lot (raising without going down) for hours this mean the model is probably done and you should select an epoch before this infinite raising
and for choosing an epoch, choose a low point in the mel graph, g/total is just an average of mel, kl and fm
very important: don't forget to sync your log!

[simple ore]

i need a bug report logged

with all the steps done

[latent kettle]

how should i do that ?/

[simple ore]

https://github.com/IAHispano/Applio/issues

[quasi dagger]

I'd say also include the model that was causing the problem

[latent kettle]

i have to sign up

[latent kettle]

every model downloaded from weights.gg

[coral frigate]

my model is going to be on 40k so ill just have to change the config file in the v1 config folder right?

[simple ore]

at least explain the steps

[analog obsidian]

yep

[simple ore]

v2 config

not v1

[analog obsidian]

mainline has it on v1

[quasi dagger]

I think he drag and dropped the zip in that drop files in the download tab of Applio

[simple ore]

whhahhha

[coral frigate]

there is no 40k.json file in v2 folder

[analog obsidian]

[simple ore]

mainline is v1????

what the fuk

[analog obsidian]

40k? yeah

lol

v2 40k uses v1 40k settings (in mainline)

[latent kettle]

Step 1 > Download any model from Weights.gg
Step 2> Extract it anywhere. You will get Model.pth and model.index from it.
Step 3> Download Applio from official page (Applio 3.2.7)
Step 4> Run "run-applio.bat" and wait for few seconds
Step 5> Go to the Download Tab in Applio GUI
Step 6> Now select the model you have Extracted (Step >2) upload model.pth (No error)
And Model.index (error)

[simple ore]

change step 2 to 'Extract the .zip into Applio's logs/modelname folder'

skip steps 3-6

[latent kettle]

Yes it is working. But I'm just informing you that it's actually broken. Fix it in next update

It's a user feedback @simple ore

[simple ore]

it can not work with the model schedme provided by weights

it expect an actually unique model name in the file name

not a .json with all descriptions

why weights decided to use model.pth and model.index is just silly

[latent kettle]

I changed the name of index by this and it worked added_IVF738_Flat_nprobe_1_Name_v2.index

Even if I just write anything like Drjr.index it don't work it only works when I put added_IVF738_Flat_nprobe_1_drjr_v2.index

[simple ore]

you need to have _nprobe_1_ in the name, it is used as a split

yes, I guess it can be fixed

but since there's no name provided by weights, what do you even expect

if you have to rename files in order to load them into Applio, then the problem is not that

[latent kettle]

Nothing. Just nick said me to inform you.

[low shard]

ah thought u made a fix for that

maybe u could talk with Vijay about this? And find a solution?

[quasi dagger]

But then all the other models on weights would need to be updated

Also that's kinda hidden feature of Applio, I never used it myself to be honest

[simple ore]

a solution would be loading .zip file from weights, unzipping, reading json, making a properly named model folder, etc

waste of time and effort when someone can just unzip it where they need it

[quasi dagger]

I never used that Applio feature lol

[simple ore]

doable, but my priority is making training better

[quasi dagger]

Didn't even know it existed

[simple ore]

if anything, would make reading the .zip file from the logs even better option

[low shard]

[quasi dagger]

Just tell them to not use that or rename the file

[hoary bloom]

best pitch in crepe for female voices?

[nocturne mural]

https://github.com/IAHispano/Applio/commit/f9d04c82979be060055c431098c7022d575cbf28

umm

Anyway, I've even forgotten how the split works.

catshake

but well

[candid lance]

does anyone know how to make the ais not sound like the og vocals in the song

[candid lance]

like i made a ai cover using a vocaloid song

and the ai sounds a bit liek the vocaloid

[safe atlas]

Raise the search index feature ratio

[knotty moth]

read the doc: https://docs.ai-hub.wtf/essentials/how-to-make-ai-cover/

[safe atlas]

Anyone have the annoying thing in windows terminal when using RVC, where you are constantly forced to press the enter button or else it will just not progress at all? Doesn't matter what machine I'm on, doesn't matter what version of RVC, applio or not, even UVR.

I've had this problem for well over a year and it hasn't bothered me much, but I'm trying to train a model now and I cant sit here for hours pressing the same button every 10 seconds all night.

I know it's not a hardware issue or an issue with a specific version.

Hope someone can help before I lose my mind.

[unique rock]

What is the best pre train to use?

[crude flame]

Original one

[unique rock]

Friend, can you help me? A while ago I trained a model with the crepe hop at 128 as it comes in Applio and the model eliminated its breaths and when making inferences, the breaths sound very robotic but too much, any solution? 😫 the model is of a singer

[azure marsh]

Hey, na (나)! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[crude flame]

Get a bigger dataset

And don't remove breaths

[unique rock]

thank you

How can I use mainline on mobile?

[fierce dagger]

i am trying to train my first model but i dont have an NVIDIA GPU, is there another way to do this other than RVC? i have a Gigabyte Geforce RTX 3080, but RVC says its not compatible.
i have tried to do research on the matter but i cant seem to find an option i can understand fully.

[simple ore]

missing cuda toolkit?

[fierce dagger]

is this a reply to my question?

[simple ore]

yes

[fierce dagger]

oh i have no idea what that toolkit is.

[simple ore]

https://developer.nvidia.com/cuda-12-1-0-download-archive

[fierce dagger]

downloaded and installed but still not compatible, any other suggestions?

[simple ore]

so which version did you download and what the error message (screenshot would help)?

not on this screen

there's a window opens when you start it

and i meant which rvc version did you download

[fierce dagger]

bare with me, i am green as can be at this. i downloaded it from the Pinokio app. i dont see a version.

[simple ore]

https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC1006Nvidia.7z

download, unzip to somewhere other than C:\users... like C:\RVC\

[fierce dagger]

done

[simple ore]

go-web.bat and see what it says

[fierce dagger]

should i post what it says?

oh, it said Nvidia gpu 3080 found

@simple ore =Legend! thank you! it is working now.

[latent kettle]

What it is. !

[twilit phoenix]

hey can anyone help

when im using wokada my voice sounds fine, but when i transfer to discord, it sounds so horrid and giberish (it picked up like 10% of what i said too) that I cant even hear my voice

it sounds so good in wokada, is there any way i can fix the discord problem?

im using VB-Audio Cable to transfer

[low shard]

-> #🔍│help-w-okada

Don’t ask about wokada help here, this channel is for RVC

[twilit phoenix]

alright, sorry

[low shard]

it’s fine

[low shard]

You can use the cloud versions, like mainline colab

If you mean use mainline on your phone locally via termux ( https://discord.com/channels/1159260121998827560/1289538710307602554 ), it would be a oain in the ass bc u can’t upload files directly in the ui

And it’s also not much suggested to use rvc locally unless u really have to do it locally

[pine mural]

hey

now i'm using laptopp

who can tutorial me use model voice?

@low shard can u help me

[low shard]

Inference (use models) on pre-recorded audios like in ai covers, or realtime voice changer for calls?

[low shard]

what’s ur pc gpu?

[pine mural]

hmm core 2

15gb ram

[low shard]

That’s not a GPU

You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

[low shard]

.

[low shard]

yea that’s an intel core i3 11th gen cpu

[pine mural]

can use?

[low shard]

i see from your gpu 0 that’s integrated, which is very bad

[low shard]

Reply to this

[pine mural]

hmm

[low shard]

It’s 2 different programs

[pine mural]

i'm using to cover songs

[low shard]

So i need to know what u want to do

[pine mural]

[low shard]

that!s inference on pre-recorded audios

[pine mural]

hmm

how can use it to cover a music?\

[low shard]

Your pc cpu can technically do it but it will be super super slow, also you won't be able to train (make models) but use them

You can:

• Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
  • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
  • Mainline: The original RVC
• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Ilaria RVC Zero: fastest and simplest that you can get for free
  • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Applio Colab: max 4 hours, not granted, of GPU

Also ofc you got to download the music, separate the instrumentals and vocals https://docs.ai-hub.wtf/rvc/resources/vocal-isolation/ And then use rvc to inference on the vocals with the rvc model

RVC doesn’t separate it automatically for you, except weights.gg which is on cloud and uses that with other things to automate the process

[pine mural]

hmm

have an videos?

tutorial

?

@low shard

[low shard]

Nope

[pine mural]

💀

[low shard]

all yt tuts are super outdated, dont even try to follow them

but you can just read the guides, they have everything you need to know

@pine mural you can try reading the guides, if you have issues you can ask me

if you really want the easiest of the easiest, weights.gg is the only one that automatically separates song vocals and does other things for u

[pine mural]

hey listening this

[low shard]

i can’t rn

[pine mural]

Arisu Cover

bruh

[low shard]

don’t share covers here as files

[pine mural]

oh

[low shard]

Our server got deleted for copyright once

[pine mural]

sorry

[low shard]

if you ever want to share a cover, please upload to yt and share the link in #1159290752195633273

[low shard]

it’s fine, did you make that? Bc like u are asking me how to make covers lol

[agile lynx]

-kaggle

[azure marsh]

📘 Kaggle Notebooks

• Applio Notebook, by Vidal Kaggle
• Applio Notebook, by Shirou Kaggle
• Music Source Separation, by Shirou Kaggle
• UVR5 NO UI, by Eddy Kaggle
• RVC Mainline, by Hina Kaggle
• Original W-Okada's Voice Changer, Kaggle
• Modified W-Okada's Voice Changer, Kaggle
• 🆕 UVR5 UI, by Eddy, ArisDev & Nick088 Kaggle
• 🆕 RVC AI Cover Maker UI, by Shirou & ArisDev Kaggle
• 📖 How to use RVC Mainline on Kaggle by Cauthess

Note: Kaggle limits GPU usage to 30 hours per week.

More commands: -colabs, -hf, -local, -rvc, /help

[blazing solar]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[quick spindle]

start_http doesnt launch

[low shard]

-> #🔍│help-w-okada

This channel is for RVC help only

[quick spindle]

my voice very glitchy in voice channel
chank 19200
extra 16300
rtx 2060
rmvpe_onnx

[leaden pebble]

dont you understand him?

#🔍│help-w-okada

this is the channel to get help with rvc models

you have to send the message in okada help channel

[flint solar]

go to #🔍│help-w-okada

[low shard]

Go to #🔍│help-w-okada

Wokada & RVC are 2 different programs

[coral frigate]

i just finished training my model on mainline rvc. i go into my logs to get the path bui i just see a list of g and d paths

[analog obsidian]

on mainline .pth files are in the weights folder, not logs

assets > weights

[coral frigate]

yh i checked and there is nothing in the weights folder

[analog obsidian]

oh, you didn't saved those

dont worry you can convert the generator to small model weight

go to ckpt processing

scroll down

and find something related to "convert to small weight" or whatever

then convert your best generator epoch

(generator is the G file, not D)

in your case is 40k, v2, guidance 1

[coral frigate]

where is ckpt located?

[coral frigate]

oww

ok

[analog obsidian]

is named ckpt processing because that names comes from the V1 era days xD

[coral frigate]

how did i manage to not save them

[flint solar]

next time make sure to select "save small final model at each save point" to avoid all of this

[analog obsidian]

you had this disabled

be sure to always leave that on

[coral frigate]

can i skip this stuff and just go straight to the bottom?

[analog obsidian]

yes, you don't need model fusion, that setting was made for creating new voices or trying to improve models trained with very small datasets

[limber epoch]

if anyone is interested, I integrated RVC-python inside of REAPER for direct in-DAW voice cloning, which doesn't exist in any other DAW as far as I know

[ripe bough]

Hey guys, I am completely new on this dc, my goal is to upload my voice as a file and get back an audio file with another male model. (Does not really matter which - maybe just one that is trained good and free )
Could you please point me to a guide or thread in this dc where I can read about it

[knotty moth]

would be too bad for those without decent gpu since it needs realtime performance

[last gazelle]

Hay there. Im trying to seperate the voice from the background noice trough RVC WebUI but I keep getting an error

[nocturne mural]

It is a fix to the files that are causing the error you complained about, you could manually update the file or wait for a next release in which it would already be fixed

[craggy light]

Can anyone share me a link to the RVC? It keeps getting updated and I want to do a new cover

[worn tusk]

Hi everybody, once I was done training my model, instead of there being a "[insert name here]33333.pth" there is a D_233333.pth and a G233333.pth, does this mean anything?

[simple ore]

those are training weights, saved with 2333333 name instead of step value, because you chose to save only one copy of them

you did not chose to save the model as well, so no model .pth until it finishes the training (last epoch you chose)

[craggy light]

Somebody help, I don’t know how to put the Voice Model I downloaded onto Applio

Never mind, I figured it out

[tawdry salmon]

Is it possible to get the RVC webui directML version to only use shared GPU memory?

[obtuse igloo]

I cant check in the Echo, Sup1 and Sup2 boxed, why?

[blazing solar]

/colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

[midnight bison]

I got two copyright strikes for an AI Cover of Enjoy The Ride by Krewella and She Ain't You by Chris Brown! And I don't want my channel banned, so should I delete the rest of my AI Covers (which all but three (of Find Your Flame, Bury The Light & I'm Here) have the copyright restriction) on YouTube? I don't think even making the videos unlisted or private can save my channel from getting another strike.

@mellow token @steel forge

@tired kraken @hearty idol can you help?

[azure marsh]

Hey, Team_Dog_285! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[tired kraken]

Simple fix, don't upload copyrighted songs to youtube if you don't have the license.

[simple ore]

I made Viva Las Vegas cover by Joe Biden.. it got struck down in like no time at all

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

[mellow token]

don’t gotta be pinging mods

just don’t use copyrighted material on your channel and you won’t get strikes

[low shard]

You can't do anything about that

I got 2 strikes too

I deleted all of my ai covers

unlisted & privated videos are still at risk

[low shard]

istg they are doing a spree

[midnight bison]

I realized not even privating videos can save my channel and I got one more strike for an AI Cover of Hey Mama by David Guetta and thus, was terminated.

[low shard]

Yea privating is still subjected to copyright

rip bro

[brittle wing]

F

it's just sony or other companies are doing it too?

[low shard]

Sony & UMG

[brittle wing]

yea just these two

hopefully

[low shard]

hopefully

Delete ur ai covers if u care about ur channel

[brittle wing]

yep

[brittle wing]

yea i dont publish ai covers

but i make them for fun

[low shard]

mfs speedrunning manual strikes

how do u even file such big and important form twice in just 4 hours as a company 😭

[brittle wing]

idk, sony and umg are strict with these covers 😭

[flint solar]

why u keep leaving and rejoining

goddamn

😭

[brittle wing]

idk

because im getting sick and tired of talking about my beefs

im now just here to see what's going on with noobies and other stuff :p

[flint solar]

its js discord bro fuck u talmbout

[brittle wing]

yea im trying to prevent doing beefs with other people but i made them

tho im still trying to prevent to do my tendency of beefing with others like i did with lyery and simplcup

i honestly don't know what's happening to me, it's because i'm autistic or i'm very stupid

[warm swan]

Just updated Replay and it told me there has been an AI Server update, downloadied and decompressed all the files and now it's just stuck at 100% or is it supposed to take 10+ minutes? on SSD btw

IT LITERALLY JUST GOT DONE AS I SENT THIS

grrr

fixed (itself)

[tranquil cliff]

Does anyone create models locally?

[vague gyro]

-help

[winter cairn]

How to convert my wav to pth?

[echo fjord]

Do anybody know working covergen for today?

[serene crown]

should i use OV2Super pretrain model for a 3:30 minute dataset?

[low shard]

It’s not a conversion automatically like .png to .jpg

you need to train your model

what’s ur pc gpu?

[low shard]

Tbh pretrains help mostly just for other languages as the OG one is only trained on English

[low shard]

wdym?

[simple ore]

ov2 has a bigger source set than og pretrain, so may help with a small dataset

[lone path]

Is there a way to turn on voice feedback? it was on before when i was using it for the first time now it's off

[pastel oak]

Monitor: headphones

[lone path]

in RVC app right?

[pastel oak]

Are you using realtime gui or wokada

[lone path]

realtime gui, is okada better?

[crude bolt]

hi.

[pastel oak]

For realtime gui voice feedback you need to go to sound settings -> playback -> right click your virtual cable, properties, enable the playback in there

Fork wokada is slightly better than realtime gui with some features that make it easier for the user

Normal wokada is worse atm

[lone path]

I'm using window 11 and i cant find playback on sound setting, i thought the app have audio feedback so i can hear myself speaking when i first launch it

[pastel oak]

It works on win11 but ok

Realtime gui has no feedback, get fork wokada then

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[pastel oak]

1st

[lone path]

can i send screenshot via dm then?

[pastel oak]

Post here

[lone path]

well i couldn't find playback in sound setting and i can't rightclick anything on it

[pastel oak]

Thats not the right sound settings

Look for sound device or similar, not in that area where you are

I forgot the shortcut wait

[lone path]

sound device took me back to the same setting page

[pastel oak]

press WIN+R, type "mmsys.cpl"

[lone path]

Uhh new problem there's no enable playback check box or anything close to it there's only Spatial Sound option that i can change to be on/off

[lunar spear]

How to make a minimal ping in rvc?

[pastel oak]

Mb, its the recording device virtual cable one

on the Listen tab

[lone path]

Got it tyvm!

[low shard]

I'm guessing you're talking about Wokada as it's realtime for calls, ask in #🔍│help-w-okada

@pastel oak why do people use this channel for wokada

[simple ore]

cuz the name is (R)realtime (V)oice (C)hanger 🤡

what do you expect?

[blazing solar]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[modern surge]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[lean wagon]

E:\RVC1006Nvidia>runtime\python.exe gui_v1.py
Traceback (most recent call last):
File "E:\RVC1006Nvidia\gui_v1.py", line 59, in <module>
import librosa
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\librosa_init_.py", line 208, in <module>
from .cache import cache
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\librosa_cache.py", line 6, in <module>
from joblib import Memory
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib_init.py", line 113, in <module>
from .memory import Memory, MemorizedResult, register_store_backend
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib\memory.py", line 32, in <module>
from ._store_backends import StoreBackendBase, FileSystemStoreBackend
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib_store_backends.py", line 15, in <module>
from .backports import concurrency_safe_rename
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib\backports.py", line 22, in <module>
import distutils # noqa
^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'distutils'

E:\RVC1006Nvidia>pause
Press any key to continue . . .

Any help here

[simple ore]

weird

it should be under runtine/lib?

[hallow thistle]

I see that folder name, the original RVC fork program is outdated. So I'm not sure why you're still using that when there are plenty forks of RVC available recently.

[lean wagon]

Do you have a link for what i need to get. I was provided this link by a mate that had it working

[hallow thistle]

-rvc

[azure marsh]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[idle bronze]

what are the most recomended settings

[hallow thistle]

Applio is a fork of RVC. It is recently developed, and being one of the only AI Hub recommended fork programs.

[idle bronze]

where do i get it

[hallow thistle]

Can you be more specific on what you're looking for?

[idle bronze]

im lost at onnx stuff

and advanced settings

every setting i do is shitty

what is the best setting

for rvc

[lean wagon]

So i delete the RVC thing and install Applio

[lean wagon]

Traceback (most recent call last):
File "E:\Applio-3.2.7\app.py", line 1, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'

Press any key to continue . . .

after instellation

[hallow thistle]

No, which one you've downloaded?

[lean wagon]

https://docs.applio.org/applio

https://github.com/IAHispano/Applio/releases

[hallow thistle]

Have you downloaded the compiled zip one?

[lean wagon]

This 1

With a installer

That runs a ton of stuf

It was not done aparently it got stuck so its finishing up now

[lean wagon]

Got this far now. But it seems it cant find any of my mics etc

[simple ore]

you know it is not 'realtime voice changer', right?

if you're trying inference from microphone, that maybe issue with the browser prmissions

[lean wagon]

My friend installed the RVC from github and he can use it real time

[hallow thistle]

That is not W-Okada the realtime voice changer. It is Applio the RVC audio converter.

[lean wagon]

I'm confused..
My friend send me this = https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/releases
He uses it for real time with the voice-models from this discord

[hallow thistle]

Huh? I don't know which fork your friend has used, but W-Okada is a program where you can inference straight from your microphone in realtime.

[knotty moth]

you better look for voice changer here #🔍│help-w-okada message

[lean wagon]

Thats pritty much what im looking for a input and output thingey so i can input with my mic and it outputs with the voice overlay

[hallow thistle]

This "RVC" program is an AI audio conversion from a file, not a realtime one.

If I remember, there was a .py file for real-time voice changer in original RVC? But I don't think this particular fork program would work that good.

[lean wagon]

Hm.. Is it possible to get this to work somehow cause then i have the same version of RVC as him and that works real time with Cabel output plugin E:\RVC1006Nvidia>runtime\python.exe gui_v1.py
Traceback (most recent call last):
File "E:\RVC1006Nvidia\gui_v1.py", line 59, in <module>
import librosa
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\librosa_init_.py", line 208, in <module>
from .cache import cache
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\librosa_cache.py", line 6, in <module>
from joblib import Memory
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib_init.py", line 113, in <module>
from .memory import Memory, MemorizedResult, register_store_backend
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib\memory.py", line 32, in <module>
from ._store_backends import StoreBackendBase, FileSystemStoreBackend
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib_store_backends.py", line 15, in <module>
from .backports import concurrency_safe_rename
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib\backports.py", line 22, in <module>
import distutils # noqa
^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'distutils'

[hallow thistle]

Please let me have time to respond.

Now if you got a GPU that's faster and newer than NVIDIA GTX 1000 series in your PC, you can do one on your PC.

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[lean wagon]

RTX 3080

[hallow thistle]

The first link for the forked version of W-Okada runs best.

You can go to #🔍│help-w-okada now if you have any problem running this program.

[lean wagon]

Alright

Thanks

[hallow thistle]

You're welcome. Now you may tell your friend to use the better one.

[modest roost]

hi namari can u help me pv?

[hallow thistle]

What can I help you with?

[modest roost]

I want to make a voice model, I already have the dataset and I know how to use Google Colab, I wanted to ask about the best options for the voice I want to train

[terse bone]

2024-12-07 03:30:23.090189: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2024-12-07 03:30:23.122018: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2024-12-07 03:30:23.131710: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2024-12-07 03:30:23.153429: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-12-07 03:30:24.926543: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
2024-12-07 03:30:26 | INFO | fairseq.tasks.text_to_speech | Please install tensorboardX: pip install tensorboardX
2024-12-07 03:30:27 | INFO | faiss.loader | Loading faiss with AVX2 support.
2024-12-07 03:30:27 | INFO | faiss.loader | Successfully loaded faiss with AVX2 support.
Traceback (most recent call last):
  File "/content/AICoverGen-NO-UI-en/src/main.py", line 344, in <module>
    raise Exception(f'The folder {os.path.join(rvc_models_dir, rvc_dirname)} does not exist.')
Exception: The folder /content/AICoverGen-NO-UI-en/rvc_models/teacher does not exist.
1

help

[hallow thistle]

The program asks you for a pre-trained model to be trained on?

Sorry, I don't accept friend request from a random person, so I can avoid getting filled by spam.

[modest roost]

It asks me to select if I want original, what type, and I wanted to know what batch size is suitable for my dataset

[hallow thistle]

-epoch

[modest roost]

what

[hallow thistle]

-rvc

[azure marsh]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[hallow thistle]

I was trying to bring up a guide, if there any information about it.

[sonic agate]

-kaggle

[azure marsh]

📘 Kaggle Notebooks

• Applio Notebook, by Vidal Kaggle
• Applio Notebook, by Shirou Kaggle
• Music Source Separation, by Shirou Kaggle
• UVR5 NO UI, by Eddy Kaggle
• RVC Mainline, by Hina Kaggle
• Original W-Okada's Voice Changer, Kaggle
• Modified W-Okada's Voice Changer, Kaggle
• 🆕 UVR5 UI, by Eddy, ArisDev & Nick088 Kaggle
• 🆕 RVC AI Cover Maker UI, by Shirou & ArisDev Kaggle
• 📖 How to use RVC Mainline on Kaggle by Cauthess

Note: Kaggle limits GPU usage to 30 hours per week.

More commands: -colabs, -hf, -local, -rvc, /help

[modest roost]

People, how many epochs and what is recommended for a 1-minute dataset?

[azure marsh]

Epoch: The number of iterations performed to complete one full cycle of the dataset during training. It's not possible to say precisely how many epochs you need for your dataset, you need to monitor the TensorBoard Graph to know if your model is overtraining.

[hallow thistle]

Thank you, Automazing Mode.

[modest roost]

how can i monitor em tensor board? have this on rvc disconected colab?

[surreal spruce]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[hasty kettle]

is there any recomended text to voice? who can mimic the anime sound

[hallow thistle]

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[hallow thistle]

There's one at "RVC TTS". I don't know, I've never used it.

[low shard]

There are different Text To Speech (TTS) AIs:

GPT So Vits: RVC isn't as good as GPT So Vits for tts, but gpt so vits (few shot tts, which means needs just a lil training for models) can't use rvc models (and viceversa), and its only limited to: english, chinese & japanese, if you wanna check gpt so vits instead, read https://docs.ai-hub.wtf/tts/gpt-sovits/

Freemium 11labs: A easy way to do TTS is https://elevenlabs.io/, you can't use RVC model on this but its a mostly premium easy way for good quality TTS

FishSpeech: FishSpeech is a 0 shot (no explicit training needed) TTS, if you got a good pc you can use it locally else use their site

With RVC Models:

RVC is natively for Speech To Speech, but forks such as ilaria rvc mainline & applio have built in tts (using Microsoft Edge TTS to make a generated tts audio, which i suggest you to choose a tts model that is the same gender and language of the rvc model you wanna use, and then convert it with rvc)

If you wanna do tts locally with RVC Voice Models (if you got a good pc):

• You can get Applio in our docs
• While Ilaria RVC Mainline here (no guide as of right now)

If you don't got a good pc you can do tts with RVC Voice Models on cloud:

• Ilaria RVC Zero (Running on A100 GPU, free fasted rvc on cloud) and the guide

• Use Applio UI Colab (with google colab T4 free daily limit gpu)

• if you don't wanna use edge tts, you could try another tts ai from our tts index and use the output as an input in rvc

[low shard]

the tensorboard is automatically opened on the colab page on rvc disconnected when training

check https://docs.ai-hub.wtf/rvc/cloud/rvc-disconnected/#5-begin-training & https://docs.ai-hub.wtf/rvc/resources/epochs-tensorboard/

@crude flame the “TB” hyperlink in the first link i sent is broken btw

[brittle wing]

Do pretrains help speed up model training?

[hallow thistle]

Depends on which pre-trained model you use for your training. One is fast but can give the lower quality, while one is slow but can give the higher quality, or both.

[analog obsidian]

This not how pretrains work in rvc

[analog obsidian]

Yes, in AI using a pretrain speeds up training
But every pretrain has the same training speed, in rvc

[brittle wing]

Cause the guides says pretrains help finish up a module faster.

[brittle wing]

Where did you get that from?

[analog obsidian]

Technically it does since if you dont train with a pretrain your model is going to take days/weeks to be trained

but every pretrain in rvc has the same training speed

[brittle wing]

That's nice and helpful

[hallow thistle]

Pre-trained model speed training applies for Stable Diffusion and certain LLM models, but I'm not sure about RVC one.

[brittle wing]

Could be

[analog obsidian]

They do

But the time per epoch is the same in every rvc pretrain

So they all take a similar time to cook a model

[brittle wing]

I noticed

[knotty moth]

• "faster" is not the correct term (not the time per epoch), it should be "taking less epochs" to get optimal one
• without pretrain, you could never get decent enough normal single speaker model
• the base default pretrain and some custom finetuned pretrains may have quality difference

[analog obsidian]

tldr; time per epoch is the same but you get an optimal model in less epochs

[brittle wing]

Nice thanks

It also trains faster but ok

[simple ore]

well, a pretrain does make training faster overall

with OG pretrain you may reach the target quality later than with a good one

[brittle wing]

Ok

[simple ore]

so it is not a per epoch speed, it is more like the total training duration

[fathom whale]

If i train a model on lets say very weird 2k words such as flabbergasted, skibidi etc would it have trouble with other words and vice versa

and what will happen if i train a model on like 10+ hours of speech would it be very good, no different from 30min or worse than 30min

[worldly oasis]

How to make a voice model for AI cover

please

[warm swan]

Replay has been stuck on "Starting server" for like 30 mins someone know why or how to fix it? maybe on how to start it manually?

starting "replay-server-win.exe" does nothing

[worldly oasis]

How to make a AI Voice

[cunning field]

guys some one knows how i can resume training in colab?

[brittle wing]

I know how but I can answer tomorrow cause I'm going to fall asleep

[formal patrol]

How to make ai cover can you suggest me to how to create ai cover have any youtube Video please suggest me

[ornate dagger]

Does lossless scaling program help AI voice programs perform better?

[azure marsh]

Hey, 🌙 𝒞𝓇𝓎𝓈𝓉𝒶𝓁 🌙! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[mighty river]

Which is more good to use for vocal extract INST-VOC-Mel-Roformer or BS Roformer-Viper-X?

[knotty moat]

Hi guys I’m running mangio rvc on macOS, I tried to convert a file with vocal changer but after 5 seconds it give me error

[flint solar]

Bs roformer

[hallow thistle]

Mangio as a fork of RVC GUI is outdated, just letting you know.

-gui

[azure marsh]

https://cdn.discordapp.com/attachments/1122285248844144733/1203460490475343953/caption.gif?ex=65d12cec&is=65beb7ec&hm=bd2fb8d010006dd7c6e3c1c67d3ae846fd1478e1a3124c544c31b43086fe54aa&

[hallow thistle]

-rvc

[azure marsh]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[hallow thistle]

Applio is a recently developed RVC GUI fork, and it's one of the only RVC forks AI Hub recommended.

[hallow thistle]

I'm not sure what you mean by the "lossless scaling program". I've only heard of this term being used with image or audio file encoding type.

[tame pawn]

mmvc keeps closing second it opens

need a fix.

pronto

[hallow thistle]

Huh. Which W-Okada version have you downloaded?

[tame pawn]

so its an old version

and I need to download the newest

correct?

[hallow thistle]

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[tame pawn]

send me a link.

nevermind

So ill download the newest version and it will work right.

[hallow thistle]

Pick the first link for the forked version of W-Okada. This one runs best.

[tame pawn]

where can I find that

fork is the best?

whats "forked"

[hallow thistle]

Can you tell me what GPU your PC have again?

[tame pawn]

give me the one that runs better >

all you need to know

[hallow thistle]

Please give me more time to respond.

[tame pawn]

I am

Im just responding to you faster

ur good

[hallow thistle]

No need to respond me that faster, I'm trying to reply.

[tame pawn]

Should I get the modified version of W Okada

[hallow thistle]

Uh. This forked version of W-Okada is recently developed from the original W-Okada.

[tame pawn]

so I should get the modified

[hallow thistle]

Yes.

[tame pawn]

Is that the "deiteris' modified"

or nah

[hallow thistle]

YES.

[tame pawn]

k

whats the best tts for female voice

sting operation style

[hallow thistle]

Now can you tell me what GPU your PC have again?

[tame pawn]

whats the best tts for female voices

[hallow thistle]

I'm not sure if you're being sarcasm or not, but like please understand that I don't know which voice model is the best.

[tame pawn]

so you want me to tell you my gpu and pc so you give me an optimal model.

[hallow thistle]

I wanna make sure if your PC has a GPU that's newer and faster than GTX 1000.

[knotty moth]

❌ means he doesn't have a gpu?

[hallow thistle]

I don't know. He just didn't even answer about it.

[tame pawn]

gtx 630

now what

[knotty moth]

yea not usable

[hallow thistle]

It looks like this GPU is too old now.

[tame pawn]

give me an optimal download

[hallow thistle]

Now you have a few options. You can run W-Okada on your PC with only CPU, this one don't really run best, or you can run it on a cloud service instead.

-realtime

[azure marsh]

Interaction has expired, use the command again for a new interaction.

☁️ (Online) W-Okada - Colab Version

• How to use Original W-Okada's Voice Changer Google Colab
• Modified W-Okada's Voice Changer Google Colab
• Original W-Okada's Voice Changer Kaggle
• Modified W-Okada's Voice Changer Kaggle

⚠️ Notes

• Colab free plan GPUs tipically works for about 4 hours each day
• Kaggle restricts GPU usage to 30 hours per week
• These options may not work on mobile devices due to the lack of a Voice Audio Cable (VAC)

[tame pawn]

so now what

Should i get the modified still

or nah

[hallow thistle]

Yes. I already answered it damn.

[tame pawn]

k

do i get the linux one

or the windows one

( im on windows )

[hallow thistle]

You might wanna run W-Okada on Google Colab if you feel like it's not that fast on your PC.

Depends on your PC hardware specs. Linux and Windows work roughly the same for this program.

[tame pawn]

gotcha

its downloading!

[hallow thistle]

Oh wait. This channel here is #✨│ai-help, not #🔍│help-w-okada. If you have any problem running this realtime voice changer program, you can go there at #🔍│help-w-okada.

[unique rock]

Does the new KLM 4.3 also work for singing?

[tame pawn]

yes na

it does

[hallow thistle]

I don't know anything about this particular pre-trained voice model. Better wait for a professional to give you an answer.

Please use #🔍│help-w-okada for more information about using W-Okada. This channel is about RVC the AI audio converter.

Can you calm the down? Holy shit.

[dapper grail]

Where do i download the voice changer

[flint solar]

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[flint solar]

go to #🔍│help-w-okada if u need help

this channel is for rvc

goddamn read

[dapper grail]

ty

[hallow thistle]

A RVC fork named Gyatt.

[low shard]

can't be real

[distant turtle]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[modern surge]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[crystal whale]

Error
unhandledrejection
no error stack

NotAllowedError: Permission denied
no stack

@odd shalehey can you help me

[azure marsh]

Hey, ! Hazεl! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[odd shale]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[crystal whale]

oh okay

it was working

and this happened

when i tried to open it today

Error
unhandledrejection
no error stack
NotAllowedError: Permission denied
no stack

with this error

@odd shalewhat do i do?

[odd shale]

I'm... pretty confused, sorry.

I have no clue if that error is happening to you on Okada or RVC.

Nor what can cause it.

Maybe it's something related to your antivirus?

I'm not sure.

[crystal whale]

hmm

[crystal whale]

okay imma make an exclusion

[crystal whale]

no

[odd shale]

Then i don't know what causes that error..

[odd shale]

Oh, it's w-okada.

If you want help regarding w-okada, you should go to the #🔍│help-w-okada channel.

[crystal whale]

okay

thanks

[odd shale]

🐢 👍

[crystal whale]

can you help me tho?
in #🔍│help-w-okada

[odd shale]

Maybe Shad can help you, he's more experienced on W-Okada than me.

[eager rivet]

okay @low shard these are the 2 files that were in the google drive the guy who made the rvc sent me

[low shard]

Alr u got both the pth and added index which is good

You can read https://docs.ai-hub.wtf/rvc/local/applio/#inference- where it says how to import the model manually as you got the files

[eager rivet]

presh bro thanks for all the help man

[low shard]

Yw, let me know

[unique rock]

Who has tried KLM 4.3? Is it good?

[mighty river]

Best model for removing reverb?

[odd shale]

Mel Dereverb V2 by anwuev

[odd shale]

I haven't tested it yet.

[knotty moth]

note: that's actually the normal one, and another "v1" one is named as "less aggressive"

[knotty moth]

#1315222227922452491 message

[karmic kite]

idk if this is allowed to send here but I really just need help making the model, I already have the sample audio

#1315575757338054727 message

[hallow thistle]

If you have any problem on train a voice model, go to #📑│making-models for more information.

[karmic kite]

I found one dw

sorry abt that

[hallow thistle]

The #✨│ai-help is for about RVC program.

[ornate narwhal]

Help please. In AI VOICE when I select a voice and start speaking it just mumbles

[low shard]

https://cdn.discordapp.com/attachments/1159290193619189821/1283652791931506759/sponge-bob-night_.gif

don’t use voice.ai

what’s ur pc gpu

[ornate narwhal]

what should I use?

rtx3060ti

[low shard]

Wokada (program that uses rvc, retrieval-based-voice-conversion speech to speech, models in realtime for calls)

especially the deiteris fork (fork = modified version) which gives better performance

[low shard]

good

-rt

[azure marsh]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[low shard]

1st link, deiteris fork

and yes there’s no yt tuts, all of thos are outdated

and this is the wrong channel for asking help about this, use #🔍│help-w-okada

[ornate narwhal]

Ok thx

[low shard]

yw

[eager flare]

If I have an isolated vocal that's way off the voice I want to convert onto it, should I edit the isolated vocal first to make it closer to the intended end voice? For example change the key in bandcamp first. Or is it better to just edit after.

[tame mica]

it should just be better if you editer it firsgt rather than after imo :p

[opal kelp]

-colab

[azure marsh]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[heady halo]

-local

[azure marsh]

Not available yet

[heady halo]

-rvc

[azure marsh]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[weary crane]

having this error

AssertionError Traceback (most recent call last)
<ipython-input-61-e02f5dae78ef> in <cell line: 9>()
7
8
----> 9 assert 'pretrain_type' in locals(), "You need to download a pretrain! Please run the "Download Pretrained Model" cell before continuing."
10
11

AssertionError: You need to download a pretrain! Please run the "Download Pretrained Model" cell before continuing.

i'm not trying to load a pretrain, i'm trying to begin training

[odd shale]

Which colab/kaggle notebook are you using?

And what's your pc gpu?

[weary crane]

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS?authuser=1#scrollTo=GnZdaTi6FgHf

3060

[odd shale]

3060 or 3060 ti?

It seems you're able to install and use RVC locally on your pc.

Here, there you got the docs.

https://docs.ai-hub.wtf/rvc/local/mainline

[hallow thistle]

Why go for Google Colab when you got that very fast GPU in your PC?

[hasty timber]

hey! I'm new here, can someone help me with something?

[azure marsh]

Hey, Chlowww! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[hallow thistle]

What can we help about RVC today?

[hasty timber]

can I pm you?

[hallow thistle]

If there's nothing private, no.

[hasty timber]

okay, so basically I tried uploading a voice I've downloaded but when I upload the file it says I haven't upload anything

or the file hasn't been uploaded

[hallow thistle]

Which RVC GUI fork program you've downloaded? Because there are plenty of them available.

[hasty timber]

RVC client

[hallow thistle]

W-Okada the real-time or an RVC audio conversion program?

[hasty timber]

yess W-Okada

[hallow thistle]

Please go to #🔍│help-w-okada for more information about W-Okada. The #✨│ai-help is about RVC audio conversion program.

[hasty timber]

yeah I was about to say that lol sorry

[knotty moth]

regardless if 3060 laptop or desktop 12 GB, it's recommended to train locally. don't even waste your money on any cloud services.

[analog obsidian]

laptops burn when you train models

so he has to keep an eye in the temperatures

[knotty moth]

not sure if it would actually reduce the lifespan, but you can temp limit to 80-85 C

[analog obsidian]

technically it does reduce their lifespan faster but they still live long enough

laptop temps while gaming are very high yet they still have long lifespan

training is more intensive than gaming and while cpu/gpu shut down when they reach 100 C, its better to not almost reach that point for multiple hours
so always keep an eye in the temperatures, if they reach similar temps like while gaming, everything should be fine

[royal lark]

Does anyone know how to fix or where the conversation is about the bug of the Applio models for W-okada?

blank sound, it just doesn't work trained models in applio in W-okada

[steel estuary]

anyone here ?

Hello ?

can I ask you something ?

[azure marsh]

Hey, Senpai ( Grind Arc )! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[steel estuary]

!howtoask do you think splite audio into 10 second chunks put into one file is more efficient and quality than a long audio ?

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[simple ore]

10 second chunks are way too big

[steel estuary]

so not recommend ?

I have like 20 minute of pure audio

[simple ore]

have one file ready, use in-app automatic slicer

[steel estuary]

and I research that they cut into chunks 10 second to be more efficient and quality

okay then

thank you

[simple ore]

10 second chunks are too big

during training at each step the generator takes 0.36s slice from each sample in the batch, so essentially ~29/30 of your 10sec sample goes unused

[steel estuary]

I see thanks you

so what do you recommand ?

a long 20minute pure audio ?

[simple ore]

RVC/Applio has built in slicer that does okay job

[steel estuary]

okay then

thank you

oh they program to work like that I see

https://tenor.com/view/furina-furina-genshin-furina-hair-gif-16554785690916683359

[knotty moth]

https://tenor.com/view/furina-embed-fail-gif-10775077421402251353

[hallow thistle]

https://tenor.com/view/genshin-impact-paimon-gif-18640817

[silent haven]

• rvc

[azure marsh]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[silent haven]

rvc

how big of a delay is there with the realt time vc

is there like a video i can watch lol

[azure marsh]

Hey, Nanichan! Please use the command !howtoask to increase your chance of getting help by structuring your question in a way others can understand better. Also make sure you're asking in the right help channel:

• General RVC help: #✨│ai-help
• W-Okada / Realtime RVC: #🔍│help-w-okada
• AI image related: #🔍│help-ai-art

[silent haven]

Hey I’m looking for the same thing but not a girl voice lol

[rough haven]

I need a realistic female voice, I want to gut someone, but it doesn't work because people hear AI in the voice, yesterday the admin called for a check, is there any very realistic female voice indistinguishable from the real one?

[flint solar]

look in #1175430844685484042

if u cant find one u like u can request one in #1159289738314919936

otherwise

make ur own by following the guides on how to train

[flint solar]

go to #🔍│help-w-okada bro goddamn cant u read

[fresh tinsel]

help

hello]

[low shard]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[flint solar]

😭😭😭😭

[low shard]

😭

[brisk horizon]

how do i fix the ai from breaking when singing high notes
when i do ai covers

[flint solar]

who are u using to infer?

[brisk horizon]

what do you mean

what model am i using?

[flint solar]

no

what version of rvc are u using to start with

[brisk horizon]

oh uhhh

its a platform on huggingface

llaria rvc idk what version it is

[flint solar]

ohh okay

[brisk horizon]

rmvpe+

[flint solar]

the issue is from the model ur using

[flint solar]

wait

use rmvpe

[brisk horizon]

why is that

[flint solar]

the normal one

[brisk horizon]

my model is high quality with hight quality flac isolated vocals

no reverb or any modifications

[brisk horizon]

i did but when it tries to sing high notes the model just breaks lmao

[flint solar]

model quality doesnt have anything to do with range

[brisk horizon]

it cant sing that high note without going robotic

[flint solar]

ur model wasnt trained on high pitched vocals

[brisk horizon]

shit

[flint solar]

as simple as that

[brisk horizon]

can i fix it?

[flint solar]

by retraining the model? yes

[brisk horizon]

yea on where

[flint solar]

-rvc

[azure marsh]

Suggestions for @brisk horizon

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[brisk horizon]

i made this model like back in 2022

[flint solar]

??

[brisk horizon]

would retraining the model actually work?

how do make sure it would sing higher range

[flint solar]

if u still got the dataset u used sure

[flint solar]

include high pitched vocals in the dataset

[brisk horizon]

the flac files?

[flint solar]

yes

[brisk horizon]

high pitched?

i manually pitch the vocals????

[flint solar]

no

[brisk horizon]

ok

[flint solar]

find high vocals of ur target voice

[brisk horizon]

ummm

i dont think there is

[flint solar]

then its not possible to fix

as far as i know

[brisk horizon]

fuuuuck

is there a trick around this

[flint solar]

no not really

it cant predict how itll sound like

[brisk horizon]

damn

[brisk horizon]

would lowering the semitones of vocals im trying to convert work?

or at least help

[knotty moth]

testing on a model trained using KLM 4.3
#1315222227922452491 message

[flint solar]

but ur output will sound off key

[brisk horizon]

i can rekey the instruments

[brisk horizon]

what does that do

i train models on it?

[flint solar]

klm is experimental

[knotty moth]

literally testing on high pitch vocals, tbh for most talking cases the default pretrain is fine

[flint solar]

this dude here said he trained the model in 2022

[low shard]

what's wrong with rmvpe+ btw?

from what I know:
Different types of RMVPE Pitch Extraction:

• rmvpe: A Robust Model for Vocal Pitch Estimation in Polyphonic Music, the normal version of the best pitch extraction, it's robust and not sensible to noise
• rmvpe+: has a pitch threshold, it limits the max and minimum pitch possible, basically deleting f0 values below and above certain thresholds
• rmvpe-gpu: Training ONLY, uses your gpu for the feature extraction process, using more gpu so making training faster
• rmvpe-onnx: Wokada ONLY, its a must for AMD Users who use ONNX models

[flint solar]

i just assumed that he messed with the parameters

[low shard]

ah alr

was asking in case i got smt wrong

[brisk horizon]

how do i use this

just to test if the vocals come out any better

nvm thats a model lmao

[long rose]

hey guys, I havnt done antything with AI vocals for about a year and things have changed. I cant seem to find the links to the page where i upload both a voice model and my vocal take and turn my voice to the uploaded model. Does anyone know how this is done nowadays?

i used to use both a huggingface and google colab before

[low shard]

What's your pc gpu

it's better to know if u can do it locally first

[long rose]

one sec

im using the Apple M1 chip

i cant seem to find my GPU

[long rose]

its not great compared to todays Apple chips but its good

[low shard]

oh damn

U can technically inference (use models) on that but it's not really the best

u can't even train (make models)

[low shard]

You could technically do it locally but I would suggest you cloud

[long rose]

thats okay, im using models that are already made, Im just using it for the purpose of converting vocals

how would i go about doing it cloud

is it similar to how it was in the past with the google colab space where i just upload vocals and a model to convert them into?

[low shard]

For Inference (use models) Mac, You can:

• Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides, probably won't be able to train, make models):
  • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
  • Mainline: The original RVC
• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Ilaria RVC Zero: fastest and simplest that you can get for free
  • Applio Colab: max 4 hours, not granted, of GPU

[long rose]

Excellent youve been a great help! thanks!

[low shard]

You're welcome

[cursive salmon]

is it normal to have the file size be the same when training more epochs? (one has 150 and the other has 200)
using the kaggle method

[dawn mica]

what do i do if the client doesnt load

[low shard]

if you're talking about wokada, use #🔍│help-w-okada

[simple ore]

the file sizes is always the same as it contains the fixed set of parameters

[cursive salmon]

ah okay thank u

[olive cliff]

Question: I'm thinking of training on a Lightning.AI notebook and I want to ask if the Applio notebook is version 3.28.0?

[spice arch]

How do you run the https://huggingface.co/QuickWick/Music-AI-Voices ?

[hollow dome]

Is there anything I can do if my RVC model doesn't have an index file?

[latent kettle]

Use ilaria rvc or applio

[spice arch]

Could you send the link to those please?

[latent kettle]

Which gpu do you have

[pallid umbra]

guys how to use applio

ive just installed it

nvm figured it out

[olive cliff]

Here.
#🤖│bots message

[low shard]

what’s ur pc gpu

[low shard]

you can use it without one, but the index is the accent

u should have it,

if its ur trained model, be sure to train the index

[hallow thistle]

Inference an audio with RVC model without an .index file can lead to some funny unexpected audio result.

[spice arch]

I have a m1 chip mac atm..

[latent kettle]

You can use Replay to convert vocals.

Or maybe cloud services may help you

[spice arch]

Mm whats that?

[low shard]

For Inference (use models) Mac, You can:

• Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides, probably won't be able to train, make models):
  • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
  • Mainline: The original RVC
• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Ilaria RVC Zero: fastest and simplest that you can get for free
  • Applio Colab: max 4 hours, not granted, of GPU

[cyan mountain]

Can anyone tell me how to install a specific model on Applio?

[hollow dome]

Thanks!
I'm used to doing this on colab but the colab I used went down and a lot of the huggingface models give me errors without index files - do you know of any of those that'll let me inference without one or should I just try to retrain it?

[low shard]

You should deffo train your index if you can

How did you train it btw?

[lavish lintel]

Congratulations Nick088 [ITA/ENG] by Weights!

Your Charizard is now level 46!

New move!

Your Charizard can now learn Fire Spin!

‎

‎

‎

‎

[low shard]

what's your pc gpu?

[low shard]

what's your pc gpu? to check if it's good enough for inference

[cyan mountain]

GTX 1660S

I can have a loss in time, then there will be another GPU

[low shard]

Your GPU is good enough to do inference (use models) locally (on ur pc), you won't be able to train (make models) but use them

You can:

• Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
  • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
  • Mainline: The original RVC
• Cloud (remote good pc, easier and faster than ur PC but it's limited):
  • Ilaria RVC Zero: fastest and simplest that you can get for free
  • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
  • Applio Colab: max 4 hours, not granted, of GPU

You can follow this to know how to upload models on Applio

I gave you also all the ways to run RVC, both locally & on cloud

[cyan mountain]

thank you, I will study and try to do something

[low shard]

yw

[cursive salmon]

what settings should i use for labeling sounds in audacity?

[golden yew]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[spring turtle]

-help

[golden yew]

!howtoask After Installing A New Voice Model, My Native Client Has Gone White And The Command System Has This Text "Wait Web Server.....310, And Increases 10 Every 10ish Seconds
[VCCLient] Wait Web Server....340 http