I'd suggest you to use Applio UI Kaggle or Applio UI Lightning.ai
#✨│ai-help
low shard
earnest vigil
i was trying applio but got stuck
low shard
i was trying applio but got stuck
which exact applio guide? and what's the issue?
earnest vigil
the applio collab
the tutorial says to get the link
but it doesnt appear to me
i installed the applio into the google drive
i put my dataset on the google drive
i just get a yellow interface saying i dont have any datasets
knotty moth
either you haven't run the installation cell or it actually failed by apparently "finishing" too quickly
earnest vigil
ill try reinstalating then
i think it worked
ill follow the guide now
also im doing this to a non english voice, i need to change anything @knotty moth
wild garnet
Hey everyone, I’m having trouble getting MMVCServerSIO to run properly on Windows with GPU (RTX 4060) 16GB RAM (u can ask if u need other specs but I dont think that's the problem). I downloaded the version MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a.zip and tried launching start_http.bat
C:\AI\VoiceChanger\MMVCServerSIO>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
But the window closes immediately after launch. I installed the latest Visual C++ Redistributable (x64), tried different version 17b — same result.
Any ideas what might be causing this? Is there a more stable version I should try, or something wrong with my config? Thanks in advance!
I installed vcclient_win_cuda_2.1.4-alpha.zip instead and it runs with no problem (actually I dont even know the real difference between vcc and MMVC but in the tutorial i was following MMVC was used, that's why i wanted to try that first
viral mason
You're using outdated softwareeeee
Don't get your stuff from YouTube tutorials
-# if you did
Since you have a really good Nvidia GPU I'd recommend you use Vonovox, it's the first guide just read up and if u need help I can help for a few hours then I disappear due to work
-rt
viral mason
@wild garnet if ya need help or have questions just ask here or me ^^
wild garnet
Since you have a really good Nvidia GPU I'd recommend you use Vonovox, it's the ...
thanks a lot, im gonna try with that
viral mason
No problem! If you don't really like how the models sound when using it I'd suggest the second best thing to use is w-Okada Deiteris fork or Tg developed fork
wild garnet
<@225956264604073984> if ya need help or have questions just ask here or me ^^
any reason why vac470lite is mentioned instead of VB-audio-cable?
viral mason
VB cable can cause weird problems like popping audio on windows but I still personally use it since it's easier to get working
wild garnet
Since you have a really good Nvidia GPU I'd recommend you use Vonovox, it's the ...
vonox is working :3 thanks
viral mason
Np! Glad to help :3
grizzled moat
Hey, im following this guide https://docs.aihub.gg/rvc/local/applio/#training but I don't see the sync graph option anywhere, anybody know why and if this will break things?
sudden violet
plush crest
anyone maybe has a sora 2 code pls i need it so bad
nova cairn
This doesnt really work with fortnite does it? Ive tried and it just gets really choppy when im playing
viral mason
turn your graphics down and see if it works
sage orbit
why is the voice incredibly chopped
viral mason
what's your gpu, and did u get the voice changer from a youtube tutorial
sage orbit
what's your gpu, and did u get the voice changer from a youtube tutorial
yea
gpu is rx 6650 xt
viral mason
gpu is rx 6650 xt
I'd suggest Wokada deiteris fork, since your voice changer is outdated by about a year
-rt
viral mason
just read the guide, it's the third one
rotund breach
worm this shits confusing
viral mason
what's confusing?
Mattex got it working easily u should ask them
I have to go to work so I'll be unavailable basically all day
sage orbit
just read the guide, it's the third one
aight thanks
granite falcon
anyone here?
im getting an error when i click convert audio
i have the voice model loaded and all
simple ore
i just get a yellow interface saying i dont have any datasets
your problem is
it means it could not start the app because, most likely, you skipped the step of installing it
granite falcon
question: is this channel dead or sum?
cuz how come every time i ask something here nobody ever answers?
bullshit
simple ore
cuz how come every time i ask something here nobody ever answers?
"im getting an error when i click convert audio"
may wanna start with showing the error
bro, learn how to ask for help
queen relic
Is there an AI like Suno AI that I can use?
granite falcon
"im getting an error when i click convert audio"
may wanna start with showing th...
What else do u want me to say exactly?
Its literally that
simple ore
Is there an AI like Suno AI that I can use?
there are few song generating open source models
granite falcon
I get an error when converting the audio
simple ore
Its literally that
it is not, look for the console output
queen relic
there are few song generating open source models
How to use open source models
granite falcon
Idk man, i just followed the steps of the guide
Its been some months since i last did one of these
simple ore
what guide? what app? start with that
granite falcon
But it worked last time
tame oracle
Is there an AI like Suno AI that I can use?
Yeah, there are AI music generators like Suno AI, but it depends if you want to run it locally on your computer or use a cloud-based service
simple ore
there are no mind readers here, you need to provide details
simple ore
okay, look in applio colab cell's output for the error
queen relic
Yeah, there are AI music generators like Suno AI, but it depends if you want to ...
whichever is easier to set up and use
simple ore
whichever is easier to set up and use
can try this https://github.com/ace-step/ACE-Step if you have enough VRAM
tame oracle
whichever is easier to set up and use
Then cloud-based is definitely the easiest. Tools like Boomy or AIVA let you start making music right in your browser without installing anything, and you don’t need a powerful PC. just remember that these have free tiers, so they might be limiting
granite falcon
I just hipe i dont run out of gpu time or whatever
queen relic
can try this https://github.com/ace-step/ACE-Step if you have enough VRAM
6 GB GDDR6 VRAM
median epoch
hello, i would just like to ask if there is any tutotorials or paper that i can read to learn about the codes about the ais. im currently trying to make something like neuro sama and just trying to have it plain and simple at first by making it take input from notepad and produce output onto notepad. I am having chatgpt guide me but its hallucinating alot. any advice for papers or tutorials i can take or read?
simple ore
6 GB GDDR6 VRAM
RIP
queen relic
😄
nocturne mural
hello, i would just like to ask if there is any tutotorials or paper that i can ...
median epoch
<https://github.com/HRNPH/AIwaifu>
i want to try build it from scratch using open source model while reading papers and tutorials. thanks for that thing but i want to make and learn not just get it
gritty barn
Hi, would anyone like to volunteer for a development project that takes less than 10 minutes? It's a quick task, and you have to use AI. It would help us a lot. We'll reward you with credits. Thank you!
queen relic
Then cloud-based is definitely the easiest. Tools like Boomy or AIVA let you sta...
I want to upload a vocal and have a song generated based on that vocal, but I can’t do this on the two websites you suggested.
nocturne mural
i want to try build it from scratch using open source model while reading papers...
You can use that code as a base and modify it however you like, see how it's structured, and look online for better 'components'. Otherwise, I guess you'll have to search for what you think you need using keywords and then put it all together into one.
median epoch
You can use that code as a base and modify it however you like, see how it's str...
ah. sure thanks. ill start from that then. ill go study the codes and use what i can... maybe i was too stuck with trying to build from scratch.
waxen axle
How are you doing the ai vocal
granite falcon
Huh?
simple ore
scroll up to the top of inference and show what you have there
simple ore
the error means you're attempting to use a model pth file that is damaged/incomplely downloaded/not actual pth
tame oracle
I want to upload a vocal and have a song generated based on that vocal, but I ca...
sorry for the delay, https://singify.fineshare.com/ you should be able to upload a vocal track and itll generate a song based off of it, anything past this and theres not really any other options
granite falcon
I literally got the model from here
waxen axle
What app is this
granite falcon
From this server
So
Idk what to tell you
@simple ore u saying i should try another model?
simple ore
<@155030383648440320> u saying i should try another model?
you can try
granite falcon
Well, i wanted to know what the problem was at least
simple ore
unpickle error is just that - the model is a .zip archive, it failed to extract it
granite falcon
Shit...
simple ore
what I mean, .pth is actually .zip
waxen axle
So we just gonna gatekeep the app
simple ore
not that you need to upload a .zip
queen relic
sorry for the delay, https://singify.fineshare.com/ you should be able to upload...
thanks ❤️
granite falcon
Yeah
I extracted it with winrar tho
So idk
And on the files names it says index and pth
Respectivally
simple ore
try opening the .pth with winrar
granite falcon
Bet
simple ore
see the log output
simple ore
you can skip the UI and use colab to upload the model into logs folder
there's gonna be content/applio/logs
granite falcon
On download model it says model downloaded succesfully
simple ore
just click refresh on UI after
granite falcon
On inference right?
simple ore
yes
simple ore
how do you upload the model?
granite falcon
I did everything right
simple ore
apparently not
simple ore
show me the steps
granite falcon
Okay so first
I connected
Then
Install applio
Till it finishes
Then start applio
Choose a method
I didnt touch here
Just left it as g radio
simple ore
i mean.. you go to 'downloads' tab, what do you do?
granite falcon
Download? U mean once im inside applio?
simple ore
yes
granite falcon
I drag rhe pth and index files
Like it says
Drop files
One then the other
It says model downloaded succesfully
granite falcon
simple ore
then again for index
simple ore
when I do that I get
but nothing in
Colab's output should say "ArnoldSchwarzenegger.pth saved in X:\Applio\logs\ArnoldSchwarzenegger"
granite falcon
granite falcon
simple ore
then refresh and pick the model from the list
granite falcon
I gotta press download model first no?
simple ore
no
simple ore
that button is for downloading .zip file from huggingface
granite falcon
I unloaddd and pick my model from the list
Im uploading the audio now
Im fucking done
What a waste of my time
Preciate u for trying to help i guess
simple ore
Preciate u for trying to help i guess
if you only need inference, it can run locally on CPU
not very fast, but serviceable
simple ore
I did not do anything special
granite falcon
U on public url too right?
simple ore
I've started colab, found the model you tried to use, uploaded the files, works fine
spring orchid
-colab
tame oracle
-realtime
hexed pagoda
What is the best vocal extraction UVR model? I use htdemucs_ft myself.
And is there any other, maybe better local or web solutions for vocal extraction?
latent kettle
What is the best vocal extraction UVR model? I use htdemucs_ft myself.
And is th...
Vocals fv4 gabox
hexed pagoda
Thanks
grizzled tree
guys can someone help me with this
thick ferry
is vonovox better than w okada/
bleak trout
Can anyone help me? When I speak, my real voice comes through first before the AI-generated one does
stoic depot
any help with okada pls
hasty sapphire
its sounds too robotic i feel
meager sonnet
What does Smart SINE, RNN Noise Reduction, and AP-BWE 48K Upscaler does
tame oracle
What does Smart SINE, RNN Noise Reduction, and AP-BWE 48K Upscaler does
“SINE” stands for Sine Waveform Vocoder.
It’s a smart vocoder enhancement method that reconstructs or improves audio waveforms by analyzing and resynthesizing the sound using clean sine-based signals.
RNN = Recurrent Neural Network — a kind of AI model good at handling sequences (like audio).
This module removes background noise and artifacts in real time.
AP-BWE = Adaptive Predictive Bandwidth Extension.
It takes lower-sample-rate audio (like 16K or 24K) and “hallucinates” missing high-frequency details to make it sound like 48 kHz studio-quality audio.
meager sonnet
“SINE” stands for Sine Waveform Vocoder.
It’s a smart vocoder enhancement metho...
Oh, thank you. Is it good to enable them all? I did enable RNN Noise Reduction, as it did significantly help my audio. As for others, idk.
analog obsidian
“SINE” stands for Sine Waveform Vocoder.
It’s a smart vocoder enhancement metho...
lol
smart sine just prevents the sine wave from inferencing noise
it's a noise gate
tame oracle
it's a noise gate
oh, well thanks for correcting me
tame oracle
Oh, thank you. Is it good to enable them all? I did enable RNN Noise Reduction, ...
well each one will add some latency, so if you're fine with a little extra delay then yeah do it
meager sonnet
Also I have a technical issue, idk if you know the solution. So far my AI Voice sounds amazing in discord. However if I launch OBS or a Video Game with built in VC, the program stops working. At first I thought maybe it was a microphone issue, but without the AI microphone works fine. I imagine the program isn't getting enough resources? Or is it a bug of somekind?
analog obsidian
AP-BWE = Adaptive Predictive Bandwidth Extension.
It takes lower-sample-rate au...
ap-bwe it's just an ai upscaler, doesn't make your audio sound studio quality, it's meant to give a fuller output
https://yxlu-0102.github.io/AP-BWE/ u can see the difference is quite subtle
tame oracle
Also I have a technical issue, idk if you know the solution. So far my AI Voice ...
and in those other apps you have the mic input as the virtual audio cable out?
tame oracle
https://yxlu-0102.github.io/AP-BWE/ u can see the difference is quite subtle
thanks
analog obsidian
RNN = Recurrent Neural Network — a kind of AI model good at handling sequences (...
also known as a denoiser
denoises the input audio so the embedder can have an easier job picking up the phonemes from the audio (potentially giving more stable outputs, the model might have less word slurring)
gpt doesnt know things about rvc sadly (hallucinates a lot or gives 2023 information, or even sometimes uses so-vits-svc training tips)
u can also gaslight it pretty easily
hasty sapphire
how u make it sound less robotic
mystic salmon
viral mason
modern cloud
What is the reason can someone help? While im talking sometimes the voice cracks and do a robotic sound
dark egret
it says to await pipeline
polar bough
low tangle
i forgot to press generate index before starting training and already got most of the way thru training. generating an index after the fact doesnt seem to produce a usable index file. should i probably start over
violet zodiac
I'm kinda on a little mission 😄 Looking for a realistic girl voice I can use for my TikTok streams. The thing is, I speak Turkish but most voice models are in English, so it gets a bit tricky sometimes 😅 Any tips on how I could fix that?
viral mason
violet zodiac
<:joe_weird:1159569750377119774>
any suggestion ?
hallow thistle
I'm kinda on a little mission 😄 Looking for a realistic girl voice I can use fo...
I can help about settings in W-Okada and things. But when you ask for "realistic girl voice", I'm not sure if this even allowed anyways.
merry turtle
my voice is cutting of can smn help
hallow thistle
my voice is cutting of can smn help
!howtoask
hallow thistle
What is your PC GPU? Did you follow any tutorial before this?
merry turtle
i looked a tut on yt
hallow thistle
Answer my question again. What is your PC GPU? I'll find a better W-Okada version if possible.
merry turtle
amd ryzen 5600
hallow thistle
amd ryzen 5600
That's CPU, not GPU. To check your PC GPU, open Task Manager, go to Performance tab, spot where GPU 0 or GPU 1 is in the left panel, and click one of these to reveal its full name in right panel.
merry turtle
oh sry amd radeon rx 6700 xt
hallow thistle
oh sry amd radeon rx 6700 xt
Try the better version from this guide. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-amd-intel-and-cpu-on-windows Let me know for issues or settings.
merry turtle
the application is not running
odd flower
Hello, what pretrain is best for my usecase based on your experience as there are a lot oh them.
15 minutes of my voice, as expressive as I could do, shouting, close mic talk, whisper and others.
(There are description for each pretrain but it would be helpful if someone who has experience would suggest, which ones to try)
Languages I am training English - Hindi
hallow thistle
No one wanted this, and this is not where you promote things.
hallow thistle
the application is not running
How is the program not running? Did you do thing right?
merry turtle
its says its been protected by windows but i run it anyways but nothing pops out and my voice is still cutting off
hallow thistle
its says its been protected by windows but i run it anyways but nothing pops out...
Try these settings in your W-Okada:
Chunk: around 70 - 100 ms, lower than this is possible if your GPU can handle
Extra: 2.7 s
GPU: AMD Radeon RX 6700 XT
F0 Det: rmvpe_onnx
Input: your microphone
Output: Line 1 (Virtual Audio Cable)
Monitor: your main speaker, in case to hear your W-Okada.
merry turtle
i think i got the wron voicechanger i cant change f0 and on extra its says big numbers
merry turtle
i cant send pictures can i dm u
hallow thistle
No.
merry turtle
could u send me maybe the voice changer
hallow thistle
You can talk a bit until your name turns blue/green, so you should be able to send an image here.
hallow thistle
This is the version I sent to you. https://github.com/deiteris/voice-changer/releases/download/b2332/voice-changer-windows-amd64-dml.zip
merry turtle
thx
hallow thistle
Also, try Virtual Audio Cable lite instead of VB-Cable, as VB-Cable gives random issues to Windows users. https://software.muzychenko.net/freeware/vac470lite.zip
hallow thistle
This is the original version of W-Okada, which is outdated and bugged, not recommended to run.
patent trellisBOT
craggy bough
now i cant hear myself in dc
merry turtle
ye i did it and still cant hear my self
hallow thistle
now i cant hear myself in dc
Input: your microphone
Output: Line 1 (Virtual Audio Cable), not CABLE Input
Monitor: your speaker/headphone
On Discord, there is this.
Also, Virtual Audio Cable and VB-Cable are two different programs made by different authors.
merry turtle
how do i install Virtual audio cable
hallow thistle
merry turtle
wait i got it
hallow thistle
You download the zip, use WinRAR or 7-Zip to extract the zip, go to the extracted folder, spot "setup64" and then double click that program.
odd flower
Hello, what pretrain is best for my usecase based on your experience as there ar...
Hey, anybody, replying to my own message so I wont lost
merry turtle
i still cant hear myself
hallow thistle
Hello, what pretrain is best for my usecase based on your experience as there ar...
In RVC, like Applio RVC fork, "original" pretrain models that come in Applio are still preferable, regardless of which voice language you want to train, compared to some other custom and third-party pretrains which can cause some issues to the voice model after training.
hallow thistle
i still cant hear myself
I think I've said "set monitor to your speaker/headphone on W-Okada" a few time already.
If you still have no idea, this could be your answer. https://cdn.discordapp.com/attachments/1425025269949010070/1426904879909306428/image.png?ex=68f037f9&is=68eee679&hm=7661842d84b8e32e492fc0120ac76208a7514f072c104a0bdcbf4e54cdce9ed7&
Anything else, you can send your whole W-Okada screenshot to here.
odd flower
In RVC, like Applio RVC fork, "original" pretrain models that come in Applio are...
okay thanks
hallow thistle
Wow, there are so many audio output devices there. How do I know if one of these is actually your headphone or speaker that you currently use?
I think the one that says "AMD High Definition Audio" might be the main speaker for sure, since its audio system looks similar to "Realtek HD Audio" as in my laptop.
Unless you plugged a headphone into a "HyperX" sound card in your PC, just test one.
merry turtle
it says that always when i start my browser what should i do here
hallow thistle
Excluding the VAC and VB-Cable input devices, the thing is how many microphone does your PC have? And which one of these is the actual functional microphone?
merry turtle
finallly
hallow thistle
Aside from the program itself, I now found some other issues that I still haven't get an answer.
merry turtle
namari ty one last question can i turn off that i can hear myself on the browser or like lower the sound
hallow thistle
<@561807329977434122> is vonovox better than w okada?
As much as other people say, Vonovox is better than W-Okada at audio quality, but its UI is a bit less friendly and more professional than other W-Okada forks. While I never test Vonovox and Deiteris/Tg Develop W-Okada forks against each other myself, I can say pretty much that.
hallow thistle
namari ty one last question can i turn off that i can hear myself on the browser...
To disable hearing yourself in W-Okada, set "monitor" to none on W-Okada.
merry turtle
my monitor and output are switched i think when i do output on none i cant here myself on browser but on dc
hallow thistle
my monitor and output are switched i think when i do output on none i cant here ...
What does this even mean? To use W-Okada with Discord or other voice chat program, you wouldn't need to set "monitor" to your speaker to hear yourself to see if the program actually working while focusing on others.
On W-Okada, set "output" to "Line 1 (Virtual Audio Cable). The "monitor" one is basically a second output.
stone dawn
Is there an AI that can take a song originally sung by a female artist and make it sound like it’s sung by a male voice — one that actually sounds masculine, not just a pitch-shifted version of the original?
https://youtu.be/cNsVMveDl8k?si=VKNZzwmPQ72VKWiS
For example, like this (they say they used an AI voice), but it has its own emotion/flow/melody, you know what I mean, instead of just pitching down the female singer’s voice and changing the tones a little.
warm peak
i just got my hands on w-okada but i genuinely have no idea what im doing
im playing around with the settings but i don't know how to get it to not sound like a load of crap
violet zodiac
I can help about settings in W-Okada and things. But when you ask for "realistic...
i send a request
hallow thistle
pale wedge
what is gpu process isnt usable??s
hallow thistle
what is gpu process isnt usable??s
What does this mean?
warm peak
what F0 detector should i use
i just downloaded this version of w-okada: MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.15
hallow thistle
what F0 detector should i use
What is your PC GPU?
warm peak
im trying to make it not sound like its going through AI
warm peak
What is your PC GPU?
Nvidia RTX 4050 6GB
i dont know if its that my GPU is just not strong enough since its not really using that much GPU anyway, like 20% max according to task manager?
hallow thistle
Download Deiteris "b2332" W-Okada instead of the one you have downloaded. The one you downloaded is the original version of W-Okada, which is outdated. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows
warm peak
oh thank you
hallow thistle
For W-Okada on NVIDIA GPU system, F0 setting is always rmvpe. For any other GPU (AMD/Intel), F0 would be rmvpe_onnx. Let me know for settings.
warm peak
okay
polar bone
yo my voicechanger sounds weird can any1 help me?
hallow thistle
!howtoask
polar bone
no i just need help with my settings
can u come in vc rq and tell me if it sounds weird and help me with my settings?
hallow thistle
When you say "catfish" at first, I'm still not sure if I should help or leave you go for whatever it is.
polar bone
uhm its not for catfish i was joking
hallow thistle
You might be joking, but some other moderators won't be playing around with that subject. 
craggy bough
-# say no :3
stone dawn
Can Ai voice be sounds real with emotions like
polar bone
hello?
warm peak
okay i just downloaded the deiteris fork how do i start it up
hallow thistle
okay i just downloaded the deiteris fork how do i start it up
Double click on MMVCServerSIO to launch the program.
In terminal window, wait until pretrain models finishing download and the program will launch your browser, which is completely normal for this W-Okada.
polar bone
please help me🙏
hallow thistle
This is your settings:
Chunk around 110 - 130 ms, lower than this is possible
Extra: 2.7 s
GPU: NVIDIA GeForce RTX 4050
F0 Det: rmvpe
Input: your microphone
Output: Line 1 (Virtual Audio Cable)
Monitor (a second output): you can set this to your speaker/headphone to hear W-Okada.
warm peak
This is your settings:
Chunk around 110 - 130 ms, lower than this is possible
Ex...
alright thanks!
hallow thistle
You're welcome. 
warm peak
okay i got it so i can hear myself now i just gotta figure out how to make it so i dont sound like im a martian
what should i put the pitch, index and formant shift thing to
hallow thistle
Any pitch number, +12 for female voice while -12 for male voice, but leave formant and index as default.
warm peak
so as in if its a female voice im trying to use i'd put it to +12 or would i do that for my voice
short hazel
🥲
rose ice
I trained a model on an NVIDIA T4 GPU using a 40-minute audio file that I downloaded from YouTube as an MP3, then converted to WAV. The training was done with a batch size of 8 for 250 epochs, but the output contains noticeable robotic noise. How can I fix this? I’m sharing the result below.
analog obsidian
this happens when you train without a pretrain, be sure you enable the "pretrained" box in the applio training ui
rose ice
Thank you!!!
knotty moth
this happens when you train without a pretrain, be sure you enable the "pretrain...
nah even with a pretrain it could still produce bad results with mediocre dataset and such
analog obsidian
nah even with a pretrain it could still produce bad results with mediocre datase...
my guy, that output sounds like an undertrained generator and discriminator
rose ice
I thought I didn’t need to enable the pretrained option since I was training a new model from scratch.
Thank you for your help!
analog obsidian
I thought I didn’t need to enable the pretrained option since I was training a n...
knotty moth
my guy, that output sounds like an undertrained generator and discriminator
the old pretrains from last year together with mediocre dataset could also produce as bad as that
analog obsidian
the old pretrains from last year together with mediocre dataset could also produ...
because they are undertrained and trained with a broken disc
knotty moth
because they are undertrained and trained with a broken disc
still I don't think training from scratch with inadequate dataset length could achieve like that, it would stay being nothing but static noise
analog obsidian
still I don't think training from scratch with inadequate dataset length could a...
then ur wrong
just do try it yourself, gather 1 hour set, batch 32, train from scratch, you'll get a similar output to that at 250e
he even said he unchecked the pretrained box
what more proof you need?
like, literally read the message above us
the generator is actually quite smart at learning your dataset
even from 0
so-vits-svc training was completely from scratch back then (which is why results weren't as good as rvc, which uses a pretrain)
knotty moth
u could have told me to give him image perms instead of having him dm the screenshot and u forwarding it here
knotty moth
I trained a model on an NVIDIA T4 GPU using a 40-minute audio file that I downlo...
I haven't read if you really said to have unchecked the Pretrained option, causing seemingly my "misunderstanding"
analog obsidian
still, by hearing that output alone you can guess the d and g are undertrained but ok... let's not continue arguing about it
honest cave
yo which one do i run guys
viral mason
so as in if its a female voice im trying to use i'd put it to +12 or would i do ...
Anything around ranges of 3,6,10,12 whatever sounds good for u
analog obsidian
yo which one do i run guys
setup64
viral mason
yo which one do i run guys
Setup64
knotty moth
yo which one do i run guys
not 64a, it's a trap
honest cave
okie ty!!
honest cave
not 64a, it's a trap
what why not 64a though
knotty moth
what why not 64a though
if you have snapdragon laptop u can try it, but it can't run voice changer
honest cave
if you have snapdragon laptop u can try it, but it can't run voice changer
oh alr
viral mason
Why does 64a even exist
analog obsidian
Why does 64a even exist
mobile devices
analog obsidian
(no, you can't use vac lite on android or iphone lol, only arm64 windows)
void raft
I have a slight problem with wakada... Monitor option does not work for me, no matter what I try simply can't hear myself. Has anyone else encountered this behavior?
analog obsidian
For what purpose would anyone need that tho on a phone
its for snapdragon laptops (idk why is this a thing lol)
viral mason
its for snapdragon laptops (idk why is this a thing lol)
Perfect for snapdragon 
viral mason
I have a slight problem with wakada... Monitor option does not work for me, no m...
You could download voicemod and just use the hear myself option as an easy fix
sterile bolt
auto closes after download
viral mason
very helpful
acoustic quartz
Yo, I tried to run this code right here: "https://colab.research.google.com/github/SociallyIneptWeeb/AICoverGen/blob/main/AICoverGen_colab.ipynb#scrollTo=NEglTq6Ya9d0", but it keeps saying "ModuleNotFoundError: No module named 'sox'", someone could pls give me a hand?
acoustic quartz
Yo, I tried to run this code right here: "https://colab.research.google.com/gith...
I'd love to send some images to illustrate, but I can't :/
low tangle
anyone know why my my g/total loss is being plotted as NaN? im using applio on colab pro (T4), batch size 4, large dataset (68 minutes, ive heard >1hr is too long for colab rvc but chat said it was fine so i thought id try it, it had no issues with a 53 minute dataset yesterday), epochs taking 7-10min. im assuming its the dataset size but im wondering if using a higher batch size could fix it
analog obsidian
anyone know why my my g/total loss is being plotted as NaN? im using applio on c...
maybe something is wrong with the enviroment/dataset
something corrupted maybe
low tangle
ill redo it and see what happens
analog obsidian
yea try doing a new session and do everything again
if that doesnt help maybe enabling fp32 would fix it but with a t4 the training will be very slow
analog obsidian
if that doesnt help maybe enabling fp32 would fix it but with a t4 the training ...
and if that still doesnt fix it, maybe its the gpu itself causing the NaNs
low tangle
thank you!
analog obsidian
dawn dagger
Hi, I need the files for the AMD GPU
serene schooner
i keep getting 'wait web server' thing and idk how to fix it, i have an rx480 8gb and windows 10. is that the reason?
dawn dagger
i keep getting 'wait web server' thing and idk how to fix it, i have an rx480 8g...
the recommended thing is to have the AMD 5000 series, I wouldn't believe I know, I have a 6GB RX 5600 xt
I don't know how it would be
tame oracle
-rt
tame oracle
@serene schooner the 3rd link would be best for you since you have a AMD GPU
serene schooner
<@1078411214578782278> the 3rd link would be best for you since you have a AMD G...
alright thanks
tame oracle
alright thanks
glad to help
lyric forge
what do i do every time i talk it keeps cutting out
bleak trout
why do i hear my own voice first and then the voice changer when i speak
knotty moth
anyone know why my my g/total loss is being plotted as NaN? im using applio on c...
- batch 4 is too low, try at least 8
- maybe dataset consistency issue
- as said, you can try fp32 though it'll be slower
low tangle
- batch 4 is too low, try at least 8
- maybe dataset consistency issue
- as said...
i switched to 6 and got the dataset under 60min and its been working normally
oh and one of my dataset tracks was stereo so i fixed that, idk if its relevant since it didnt have any issues in the preprocessing/feature extraction
analog obsidian
i switched to 6 and got the dataset under 60min and its been working normally
it's a myth rvc works bad with bigger datasets
i'd say is the opposite, is better with big datasets and bad for small datasets lol
GANs work better with more data
the NaNs were something else tho, unrelated to dataset length
low tangle
it's a myth rvc works bad with bigger datasets
oh good to know, i thought there was kind of a sweet spot
analog obsidian
oh good to know, i thought there was kind of a sweet spot
maybe 200 hours could be the max rvc can handle? idk the generator of rvc is pretty small, so it shouldnt be able to handle very big stuff (big as over 1k hours lol)
but with bigger datasets you get more stability during the training, most of the time this translates as having a less robotic model
in smaller datasets the discriminators gets too strong and thats what gives the peculiar robotic sound we all know
with more data you prevent that from happening too early during the training
so yea don't be afraid of training 1 hour datasets or more
inland yarrow
hi guys :) just wondering if ai voice changers are safe to use in CS2? (specifically w/ okada voice changer or if there are other alternatives)
viral mason
hi guys :) just wondering if ai voice changers are safe to use in CS2? (specific...
They're completely safe but may not run well unless you lower your graphics, it's also dependent on which voice changer you're using and your gpu
knotty moth
hi guys :) just wondering if ai voice changers are safe to use in CS2? (specific...
the anticheat won't flag it
inland yarrow
the anticheat won't flag it
perfect, thank u guys :) saw a steamforum post about it saying they got VAC'd so was a little worried
knotty moth
perfect, thank u guys :) saw a steamforum post about it saying they got VAC'd so...
no I didn't mean to deliver such good news, sorry that I didnt know
you could have said about the VAC ban before
at last it comes to your own responsibility
but anyway there have been some ppl using voice changer in valorant and marvel rivals
inland yarrow
you could have said about the VAC ban before
ah oops fair- the post was like 2 years ago and its like th eonly post ive ever seen on it.
shadow stream
This might not necessarily be Ai voice related, but can anyone help me create or find a voice changer that can make my voice sound like this https://www.tiktok.com/@mesaakkk/video/7557494469289954591?is_from_webapp=1&sender_device=pc&web_id=7366430486748808734
meager comet
Hey guys! I'm having a problem. Before on Windows 10, my W-okada voice changer
voice changer worked perfectly when I played heavy games like Monster Hunter Wilds, etc. Now that I've upgraded to Windows 11, when I open the game, the program cuts off the voice until I change windows. I don't know how to fix it... Can anyone help me?
lime sentinel
Hi everyone, im trying to compare raytracing and NN on calculating time and my samples are mesh with visibility on it. A node got a binary information, 1 if visibile or 0 if not. I'm trying to figure out which type of NN would be the best. Im focusing rn on GAT but do u have any other ideas ?
unreal linden
Hello, from last few days I'm getting a robotic sounds in all the voice i use.
What's the problem
void raft
Okay, this is new for me. How the fk make Wakada stop saying "trial" constantly?
Oh, looks like it is VLC. Somehow downloaded trial version
simple ore
Oh, looks like it is VLC. Somehow downloaded trial version
you've downloaded a trial version of the virtual audio cable instead of v4.70 lite
remote perch
hi guys am using okada and its working fine but its only using my cpu, i've installed onnxdirectML-cuda amd version
fallen condor
Hi guys, why does my sound break when I connect to the game?
bleak trout
why do i hear my own voice first and then the voice changer when i speak
low shard
hi guys am using okada and its working fine but its only using my cpu, i've inst...
elaborate your pc gpu and the download link, take a look at the help guidelines please
viral mason
Pretty sure whatever he downloaded is ancient technology
low shard
Hello, from last few days I'm getting a robotic sounds in all the voice i use.
uninstall vb audio cable, get vonovox with vac lite, try other models
remote perch
elaborate your pc gpu and the download link, take a look at the help guidelines ...
nvm thanks, but the voice seems chuppy sometimes like it sound robotic
girly but robotic
how i can make it smoother?
bleak trout
<@840639134091706378> <@1274496965136089149> both of yall please elaborate
My gpu is rtx 3050 and i use windows 11. Im trying to use the voice changer but whenever i speak, i hear my own voice first and then the Ai voice. I used the Vonovox Guide
void salmon
hello here. Sorry to come back from the dead, I'm just having issues. I cannot find any good AI hub to make simple ai covers. LIke everytime I try one, I put the model onto download model, everything runs fine, but it doesn't apply the model. So is there a simple one like Ilaria RVC did ? Thanks a lot 😉
torn edge
im using deiteris voice changer and using "client" audio doesnt work
it wont output anything
"server" works
but not client
simple ore
but not client
client requires browser permissions to access mic
meager comet
Hey guys! I'm having a problem. Before on Windows 10, my W-okada voice changer
voice changer worked perfectly when I played heavy games like Monster Hunter Wilds, etc. Now that I've upgraded to Windows 11, when I open the game, the program cuts off the voice until I change windows. I don't know how to fix it... Can anyone help me? Or anyone know the last okada uodate or a better voice mod?
My setup
RTX 3090
I7 13700KF
32GB DDR5 6000HZ
WINDOWS 11
simple ore
Hey guys! I'm having a problem. Before on Windows 10, my W-okada voice changer
v...
probaby HAGS
turn it off
better voice changers are here
-rt
fleet marsh
how do i use a local realtime rvc?
thick latch
Is this tool of playing a voice file reliable? I would like to use to test a certain model in different pitches but without the annoyance of having to speak with myself all the time
thick latch
For some reason I can't load any audio file there tho why
Nvm apparently it only works with WAV files despite the page saying it accepts FLAC and mp3
stray copper
whats the best UVR5 Overlap
for Vocal/Instrumental/Karaoke models
im using 5 rn
SHUT THE FUCK UP
orchid peak
I speak Portuguese and I'm using a translator, I wanted to know if anyone could help me use the aatrox voice, I managed to put it in the app but I think I did something wrong
low shard
My gpu is rtx 3050 and i use windows 11. Im trying to use the voice changer but ...
show ur audio setup
low shard
I speak Portuguese and I'm using a translator, I wanted to know if anyone could ...
hello, could u please elaborate?
!howtoask
bleak trout
show ur audio setup
I cant send Screenshots
oak schooner
Are these okay? Still feels weird when something is going up.
Codename's fork
Pretrain LegacySpinV2, batch 12, 40min data.
rose ice
Due to resource issues, my work was delayed for a while, but I upgraded my computing environment on Colab (paid plan) and finally tried building a model in earnest.
Is it normal for the output to sound like this at around 300 epochs? The pronunciation seems a bit slurred or unclear…
The file 1017orivoice is the original audio, and 1017outvoice is the one generated by the model.
I paused training at around 120 epochs and later resumed using the same G and D weights. Could that be the cause of the problem?
⸻
analog obsidian
Due to resource issues, my work was delayed for a while, but I upgraded my compu...
download v2.5 for contentvec
v2 is broken and i forgot to delete it from the huggingface repo
this
only for contentvec, dont select spin in the feature extraction step
rose ice
thaks
rose ice
only for contentvec, dont select spin in the feature extraction step
Thanks!
pure root
rtx 3060ti
ryzen 5 5600x
Windows 11
HAGS is off
im using VAC Lite and W-Okada 2.0.78 beta, I tried using the non beta version but it was much worst.
It is very choppy, words often sound unclear, glitches out halfway through.. etc. I have tried up to a 1s delay, but im also not too familiar with the software
rigid bobcat
hello I have a voice acapella that is around 20 seconds. Is it possible to train it properly?
hallow thistle
rtx 3060ti
ryzen 5 5600x
Windows 11
HAGS is off
im using VAC Lite and W-Okada 2...
Try Vonovox or Deiteris/Tg Develop's fork W-Okada. https://docs.aihub.gg/realtime-voice-changer/local/vonovox/ https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-amd-intel-and-cpu-on-windows The one you tried to run is the original W-Okada, 2.0.xx is its true latest version but still outdated.
hallow thistle
how do i use a local realtime rvc?
Which local realtime RVC? Are you looking for W-Okada or something else?
pure root
Try Vonovox or Deiteris/Tg Develop's fork W-Okada. https://docs.aihub.gg/realtim...
okay, im gonna try Vonovox
pure root
Try Vonovox or Deiteris/Tg Develop's fork W-Okada. https://docs.aihub.gg/realtim...
works wayyyy better
viral mason
Due to resource issues, my work was delayed for a while, but I upgraded my compu...
don't pay for colab just use kaggle which gives 30 hours for free users, which is more than enough to train a voice
brittle wing
does anyone have a Sora 2 invite code?
or does anyone know how to make AI vids w/out the 'policy guidelines'?
pure root
Problem im having now is that my game fps drops extreme when using Vonovox
tribal mauve
Yo guys im using applio realtime voice changer atm and was wondering if i can get better latency cuz it takes like 10 seconds and its not working properly either... is there anything i have to change ?
simple ore
Yo guys im using applio realtime voice changer atm and was wondering if i can ge...
what gpu, what latency does it show on UI?
viral mason
Problem im having now is that my game fps drops extreme when using Vonovox
lower the game settings
nocturne pivot
pure root
lower the game settings
Im normally getting like 80-100 fps gta, but with Vonovox running im at 3-5 cant even play
viral mason
Im normally getting like 80-100 fps gta, but with Vonovox running im at 3-5 cant...
I wouldn't know
I only play vrchat with the voice changer
I have a 5070 ti tho
viscid moss
@crimson depot try reinstalling UVR5 UI
Also lmk if u are going to use normal installation or precompiled
Bcuz there's no new precompiled version released yet
crimson depot
<@806611115059445830> try reinstalling UVR5 UI
Ok. One more question. Can I install the cover collab locally on my pc to run it with the 1660ti?
tribal mauve
what gpu, what latency does it show on UI?
sometimes 1000 and another time like 700-800 ggs
im cooked haha
bleak trout
Does anyone know how to make it sound more realistic? I wanna troll
viral mason
what kind..
bleak trout
what kind..
karaoke
viral mason
mk
gritty briar
How do I tran a voice model on weights.gg
simple ore
sometimes 1000 and another time like 700-800 ggs
that seems like running on CPU numbers
honest depot
i have the thing on start and everything
don't hear a thing
it says: vol 0 buf 680 ms res 16 ms rtf 0
the normal voices work but not my custom 1
mb.
Full GPU Name: NIVIDIA RTX 5060 8 GB
Operating System: windows 11
Only option that works for the ai is "Cpu", not my gpu.
viral mason
How do I tran a voice model on weights.gg
Don't, it'll come out awful
Use applio instead
gritty briar
Use applio instead
Is it an app or website
viral mason
Is it an app or website
it's able to be used both locally and on the web like on google colab or Kaggle
gritty briar
it's able to be used both locally and on the web like on google colab or Kaggle
Alright thanks
viral mason
I'm currently busy but one of the mods here would be able to help you
tribal mauve
that seems like running on CPU numbers
is this fixable or is my pc just shii ?
viral mason
is fv4 still recommended for datasets? it comes out like this at times which sucks
@sonic agate how's fv7 going btw? sorry for the ping if you're busy
simple ore
rx 580
did you follow the AMD install guide?
or you just started applio and hoped for the best?
you may need to get DML version for your RX 580, applio does not support DML model
gilded robin
hey i'm having an issue with wokada, the ai voice cuts off at the end a bit too early/abruptly and cuts off like 0.2s of speech and then the next time i speak it adds that last part to the start of my sentence
what could be reasons for this?
tg develop fork
sonic agate
<@927944475513352235> how's fv7 going btw? sorry for the ping if you're busy
blocked
jk
testing
viral mason
nocturne mural
is fv4 still recommended for datasets? it comes out like this at times which suc...
I usually use bigbeta6x + fv4, and I haven't had any problems so far.
viral mason
I usually use bigbeta6x + fv4, and I haven't had any problems so far.
never heard of that first one
nocturne mural
never heard of that first one
viral mason
numbers don't really do anything for me, I can only go off examples to compare
nocturne mural
numbers don't really do anything for me, I can only go off examples to compare
On that we agree — I just based it on the fact that BigBeta6x worked really well for me when separating vocals from an instrumental, so trying it for speech separation isn’t a bad idea either. And that’s what I did — so far, I haven’t really noticed any difference compared to FV4, except for some occasions where a few effects would slip through, but I just removed those manually.
nocturne mural
So, from my experience and for my own uses, that combination hasn’t caused me any issues. If you need proof, you can run your own tests and see which result you prefer.
nocturne mural
🐱👍
honest depot
my voices sound so robotty and trash compared to the samples
windows 11
rtx 5060 8 gb
https://www.youtube.com/watch?v=SxdnGxicJOg
viral mason
my voices sound so robotty and trash compared to the samples
windows 11
rtx 5060...
if u used a yt tutorial you have some really old software
ur gpu is so much better for vonovox
-rt
viral mason
first guide
honest depot
if u used a yt tutorial you have some really old software
i'm using wokoda, what's the best one?
honest depot
first guide
oh shi alr
viral mason
oh shi alr
if u got any issues u can always switch to wokada deiteris which is the third guide
but do ask here or ask me if u got any questions
honest depot
but do ask here or ask me if u got any questions
alright thanks bro
viral mason
np!
honest depot
oh yuh, @short fossil can it sound like a actual human?
i was thinking of making my mic like lower quality or add background noise to mask if it sounds a bit off
viral mason
u pinged the wrong person lmao
viral mason
but it should if u try yea
honest depot
aight
honest depot
but it should if u try yea
C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.30143.rartemp>runtime\python.exe launcher.py
The system cannot find the path specified.
C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.30143.rartemp>pause
Press any key to continue . . .
do i need to download python
viral mason
oh yuh, <@780574146856878100> can it sound like a actual human?
just gotta have good settings and also a good model
viral mason
C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.30143.rartemp>runtime\python.exe ...
uhhhh not sure, download it tho in case
viral mason
ye i jus needed python
cool
desert star
my voicechanger sais Pipeline not initialized how do i fix?
simple ore
do i need to download python
you need to unzip whatever you've downloaded properly first
honest depot
you need to unzip whatever you've downloaded properly first
fawk i lowk forgot ☠️
simple ore
because this 'C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.30143.rartemp>runtime\python.exe launcher.py' what you may see trying to run the file from the temp view of the archive
honest depot
because this 'C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.30143.rartemp>runti...
but it downloaded after installing python
simple ore
the project should not be in 'C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.30143.rartemp'
it is a temporary folder in a Temp folder
honest depot
do i have to reinstall after this then?
simple ore
worse place would be unzipping it into recycle bin
honest depot
the project should not be in 'C:\Users\cdcoc\AppData\Local\Temp\Rar$DIa25584.301...
that's just what popped up when i tried to install
simple ore
if you download a compiled version of vonovox, all you need is to unzip it
honest depot
if you download a compiled version of vonovox, all you need is to unzip it
which one is the actual app
found it rnvm
we been sittin on dis hto
ok got it open thx
fossil hearth
alright so what now lol
fossil hearth
let me see
viral mason
U can check in task manager, just click proformance
Should be that second button on the side
viral mason
That works too yea
fossil hearth
NVIDIA GeForce RTX 3080(10GB)
viral mason
Ooh peak
Alr u can probably use Vonovox which is currently the best, it works with 30 series and up for Nvidia only
AMD support is coming soon
fossil hearth
vonovox?
viral mason
Just read the guide for the first one
fossil hearth
how do I get that
viral mason
-rt
viral mason
Right here friend
fossil hearth
so is it a better version?
fossil hearth
alr lol let me install it
viral mason
It's currently better than all current stuff and it's still receiving updates
The only one still getting updated till this day
fossil hearth
ohhh
alr lol
wait
can u mute ur self and just look
and u can type on this chat
viral mason
Watcha need help with
fossil hearth
downloading it
viral mason
I kinda can't join VC at all rn
viral mason
All u need is the .zip
fossil hearth
ohhh im dumb
viral mason
Nah you're good
fossil hearth
Nah you're good
wait so what do I click?
fossil hearth
alr thanksss
now last questions, does it work like the other voice changer and whats the best voice?
fossil hearth
After setup is good
mb I just need to ping u lol
viral mason
now last questions, does it work like the other voice changer and whats the best...
Idk the best voice but, all I do is insert the pth and index files where they go and press start after u fix the settings up to work well for your pc
fossil hearth
alrrrr thanks bro
fossil hearth
Idk the best voice but, all I do is insert the pth and index files where they go...
I know I asked before lol is this voice changer more belivable then the other one? I havnt kept up with the updates at all
viral mason
I know I asked before lol is this voice changer more belivable then the other on...
It should be, I've heard people say it sounds better than previous stuff
I've even tested it some and it sounds really good
fossil hearth
I've even tested it some and it sounds really good
one problem I had with other is that if you scream or cough it sounds like robot, is that problem still here?
thats why I quit with voice changers and just tryed voice impressions lol
honest depot
yo worm is there a way to make it sound like its coming from a real microphone
honest depot
one problem I had with other is that if you scream or cough it sounds like robot...
gluttony subaru?
honest depot
how do u know... 😭
i lowk just had a gut feeling
fossil hearth
most underrated anime 😔
viral mason
one problem I had with other is that if you scream or cough it sounds like robot...
Yea unless the model was trained with that kinda stuff it most likely will still have trouble reproducing the sounds
Would be less robotic tho with Vonovox
fossil hearth
oh ok niceee
viral mason
like background audio n everything
It's not possible, I've seen people ask before it's completely impossible to have the AI and real life bg sound at the same time
fossil hearth
first I kind of want a realistic girl voice, and second one that sounds like a kid so I can ragebait lol
dam the thing is still downloading
honest depot
It's not possible, I've seen people ask before it's completely impossible to hav...
I mean
I guess you could have soundpad open
on a loop of background audio
no?
honest depot
first I kind of want a realistic girl voice, and second one that sounds like a k...
Ye first one never happenin bro
Shit dont work im sry
viral mason
on a loop of background audio
Yea you could
viral mason
first I kind of want a realistic girl voice, and second one that sounds like a k...
You can look for yourself here I don't use those I use like Venom and stuff like that
https://discord.com/channels/1159260121998827560/1175430844685484042
Imma go to bed now night man, if u see me online tomorrow just dm me or @ me here
I'll respond once I'm awake
robust mirage
guys can u help me, i want to try this voice changer things, i have nvidia rtc 4070 and w11, what voice changer should i use? okada or vonovox?
pure root
extreme fps drop using vonovox and VAC. like from 90-100+ fps sometimes to below 10 unplayable, whole pc stutters until i shut down Vonovox. I dont expereience this though if I dont have a game open, is this software really that intensive ?
craggy bough
holy
dull crown
What is the best local realtime voice changer for calls/games?
hallow thistle
What is the best local realtime voice changer for calls/games?
Deiteris/Tg Develop fork W-Okada or Vonovox. What is your PC GPU?
dull crown
NVIDIA
short tree
Deiteris/Tg Develop fork W-Okada or Vonovox. What is your PC GPU?
I use amd and am also looking for once lmao. W-Okada rarely seems to work. Even when it does, its very slow despite my GPU being mid-end
hallow thistle
NVIDIA
To check your GPU name, open Task Manager, go to Performance tab, spot where GPU 0 or GPU 1 is in the left panel.
short tree
I use amd and am also looking for once lmao. W-Okada rarely seems to work. Even ...
RX5700 XT
dull crown
To check your GPU name, open Task Manager, go to Performance tab, spot where GPU...
Oh, Im using RTX 4070 TI
hallow thistle
I use amd and am also looking for once lmao. W-Okada rarely seems to work. Even ...
There are different W-Okada versions. You might have been using the original and outdated W-Okada DirectML before.
Vonovox is only for NVIDIA GPU as of now, but it's possible if the creator makes one for AMD/Intel GPU.
short tree
There are different W-Okada versions. You might have been using the original and...
using the very latest, at least i believe.
Unless im wrong
stuck elbow
can someone help me
when i upload a model is just says "PermissionError: [WinError 5] Access is denied: 'model_dir\3'"
hallow thistle
using the very latest, at least i believe.
```MMVCServerSIO_win_onnxdirectML-c...
Yes, this is the one I implied. v.1.5.3.18a is the original W-Okada version.
hallow thistle
when i upload a model is just says "PermissionError: [WinError 5] Access is deni...
!howtoask
short tree
Yes, this is the one I implied. v.1.5.3.18a is the original W-Okada version.
I got that off the Repo that was listed. idk where the latest is. Unless im just blind
hallow thistle
I got that off the Repo that was listed. idk where the latest is. Unless im just...
short tree
trying to use any of the vcclient ones result in a hard error trying to boot it
short tree
Try this version. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-...
ill give it a look. thank you very much ❤️
dull crown
Should i use W-Okada or Vonovox?
hallow thistle
So I am now helping 3 people at once. 
hallow thistle
Should i use W-Okada or Vonovox?
Either Vonovox or W-Okada. Vonovox can give better audio quality but its UI is less friendly and more professional than other W-Okadas. https://docs.aihub.gg/realtime-voice-changer/local/vonovox/ https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows
Deiteris fork W-Okada, its functions overall is similar to the original W-Okada but better.
short tree
So I am now helping 3 people at once. <:cute_bigeyes:1424602508730306570>
:((
im so mad
hallow thistle
Nah, I'm not that capable of helping many members at once, I'm slow. 
hallow thistle
extreme fps drop using vonovox and VAC. like from 90-100+ fps sometimes to below...
Vonovox uses your "GPU" to process its audio, which is why you would see your game fps drops down whenever you run the game with Vonovox. It usually happens to PC with a single GPU available, but those who have 2 or more GPUs won't encounter this issue.
dull crown
Deiteris fork W-Okada, its functions overall is similar to the original W-Okada ...
I saw on the aihub doc the tg develop can use model embedder type which deiteris cant use it, is it still better using deiteris?
hallow thistle
I saw on the aihub doc the tg develop can use model embedder type which deiteris...
The Tg Develop one simply added some more features to its fork W-Okada where some features didn't exist or went missing in b2332 and earlier versions, its UI is quite unique and different, but Tg Develop fork W-Okada overall all function similar to Deiteris because it is forked from Deiteris.
hallow thistle
guys can u help me, i want to try this voice changer things, i have nvidia rtc 4...
Is that Microsoft Windows 11 or a Linux distro?
robust mirage
Is that Microsoft Windows 11 or a Linux distro?
microsoft
hallow thistle
microsoft
Either Deiteris/Tg Develop W-Okada forks or Vonovox will work. Vonovox can give better audio quality but its UI is more professional than the others. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-nvidia-on-windows https://docs.aihub.gg/realtime-voice-changer/local/vonovox/
robust mirage
can rtx4070 run vonovox smoothly tho?
hallow thistle
do i need to download python for me to be able tp upload ai voices
You didn't answer my questions. What are your PC GPU and operating system? What are you looking for and tryna do? Are you looking for W-Okada or other RVC forks?
stuck elbow
You didn't answer my questions. What are your PC GPU and operating system? What ...
oh mb My gpu is an AMD rx 6650 xt im on windows 10
im tryting to use the voice changer
hallow thistle
can rtx4070 run vonovox smoothly tho?
I never used Vonovox myself so I'm not sure.
hallow thistle
oh mb My gpu is an AMD rx 6650 xt im on windows 10
So did you follow any tutorial before this?
hallow thistle
yes
Use the better W-Okada version from this guide instead of the one you got from YouTube tutorials. https://docs.aihub.gg/realtime-voice-changer/local/deiteris-w-okada-fork/#download-amd-intel-and-cpu-on-windows
stuck elbow
Okay
dull crown
The Tg Develop one simply added some more features to its fork W-Okada where som...
ok i pick the tg develop i guess because "more" features
hallow thistle
Okay
Let me know if you run into issue or looking for settings. Microsoft Windows 10 has ended on October 14th, 2025, so you might wanna upgrate to Windows 11 if necessary.
hallow thistle
ok i pick the tg develop i guess because "more" features
That's also a good choice.
stuck elbow
Let me know if you run into issue or looking for settings. Microsoft Windows 10 ...
It worked
thank you
hallow thistle
You're welcome. 
hallow thistle
You ask me like if I know what you're looking for. 
oak edge
You ask me like if I know what you're looking for. <a:YuukaErm:12918328144463175...
broo
i mean kaggle applio is not working
oak edge
You ask me like if I know what you're looking for. <a:YuukaErm:12918328144463175...
help me out on what's goin wrong or if there was any new update
simple ore
im gettin this
seems like you did not run the main cell that installs applio
oak edge
seems like you did not run the main cell that installs applio
i ran
i had to resort to this code at last to make it work
oak edge
seems like you did not run the main cell that installs applio
!fuser -k 6969/tcp || true
!fuser -k 8077/tcp || true
!fuser -k 9876/tcp || true
!pkill -f "lt --port 8077" || true
!pkill -f "lt --port 9876" || true
# ==== IMPORTS & HELPERS ====
import os, time, shutil
from pathlib import Path
def read_last_url(file_path):
try:
txt = Path(file_path).read_text()
lines = [ln for ln in txt.splitlines() if "your url is:" in ln]
if lines:
return lines[-1].replace("your url is:","").strip()
except Exception:
pass
return "(still starting… run again to refresh)"
# ==== CD INTO YOUR REPO ====
WD_CANDIDATES = [
"/kaggle/working/program_ml", # where your repo was cloned
"/kaggle/working/Applio",
"/kaggle/working"
]
for p in WD_CANDIDATES:
if os.path.isdir(p):
os.chdir(p)
break
print("Working dir:", os.getcwd())
# ==== HARD FIX: ensure ./assets/config.json exists ====
root = Path(os.getcwd())
pkg_assets = root / "program_ml" / "assets" # source in package
root_assets = root / "assets" # where the app expects it
if pkg_assets.exists():
if root_assets.exists() and not (root_assets / "config.json").exists():
print("Found ./assets but no config.json → replacing it from program_ml/assets")
shutil.rmtree(root_assets, ignore_errors=True)
if not root_assets.exists():
print("Copying program_ml/assets → ./assets …")
shutil.copytree(pkg_assets, root_assets)
cfg = root_assets / "config.json"
print("assets/config.json exists:", cfg.exists())
if not cfg.exists():
raise SystemExit("❌ Missing ./assets/config.json even after copy. Check that program_ml/assets/config.json exists in your repo.")
# ==== Optional: add repo root to PYTHONPATH ====
os.environ["PYTHONPATH"] = f"{root}:{os.environ.get('PYTHONPATH','')}"
# ==== FILEBROWSER (9876) ====
print("▶ Starting Filebrowser on :9876 …")
os.system("filebrowser -r /kaggle -p 9876 > /dev/null 2>&1 &")
# ==== TENSORBOARD (8077) ====
print("▶ Starting TensorBoard on :8077 …")
os.makedirs("logs", exist_ok=True)
get_ipython().system_raw("tensorboard --logdir logs --port 8077 --host 0.0.0.0 > /dev/null 2>&1 &")
# ==== LOCALTUNNEL for TB + Filebrowser ====
print("▶ Installing LocalTunnel (first time may take ~30s)…")
!npm install -g localtunnel > /dev/null 2>&1
# TB tunnel
Path("t.txt").write_text("")
get_ipython().system_raw("lt --port 8077 > t.txt 2>&1 &")
# Filebrowser tunnel
Path("f.txt").write_text("")
get_ipython().system_raw("lt --port 9876 > f.txt 2>&1 &")
time.sleep(8) # give lt a moment to print URLs
tb_url = read_last_url("t.txt")
fb_url = read_last_url("f.txt")
print("\n✅ LocalTunnel links")
print("TensorBoard:", tb_url)
print("Filebrowser:", fb_url)
# ==== PICK ENTRYPOINT (prefer nested program_ml/app.py if present) ====
candidates = [
root / "app.py",
root / "webui.py",
root / "launch.py",
root / "main.py",
root / "web.py",
root / "program_ml" / "app.py",
root / "program_ml" / "webui.py",
root / "program_ml" / "main.py",
]
ENTRY = next((str(p) for p in candidates if p.exists()), None)
if ENTRY is None:
# Fallback search
for r, d, f in os.walk("."):
for name in ("app.py","webui.py","launch.py","main.py","web.py"):
if name in f:
ENTRY = os.path.join(r, name); break
if ENTRY: break
if not ENTRY:
raise SystemExit("❌ Could not find any entry file (app.py/webui.py/launch.py/main.py/web.py).")
print("\nFound entry:", ENTRY)
# ==== LAUNCH (Gradio public URL) ====
print("\n🚀 Launching with Gradio --share (watch for “Running on public URL”) …\n")
cmd = f'python "{ENTRY}" --host 0.0.0.0 --port 6969 --share'
print("CMD:", cmd, "\n")
os.system(cmd)
sonic mauve
Hello, why should I download VAC Lite (Virtual Audio Cable by Muzychenko)?
Why is it recommended over other virtual audio cables like VB Audio Cable?
simple ore
i had to resort to this code at last to make it work
that's not the install cell
simple ore
Hello, why should I download VAC Lite (Virtual Audio Cable by Muzychenko)?
Why i...
because it works
low shard
Hello, why should I download VAC Lite (Virtual Audio Cable by Muzychenko)?
Why i...
some users reported random issues with vb audio cable on windows
low shard
can rtx4070 run vonovox smoothly tho?
that's a good gpu for it, it should, but be aware that it's recommended for everyone to play with low graphics if you're going to use both voice changer and games
oak edge
that's not the install cell
🫠it's startup cell for the Applio, the install cell wasn't changed while changing to this
I also ran install cell many times there were no uses
nocturne mural
I also ran install cell many times there were no uses
I don’t think it was a good idea to run the installation cell multiple times. Noobies mentioned it because if the file app.py wasn’t found, it means the repository either didn’t get cloned at all or is possibly in the wrong path. Are you sure you’re using the code from the latest version that was released for that notebook?
oak edge
I don’t think it was a good idea to run the installation cell multiple times. No...
Yup I'm pretty sure I've tried copying two times from the link from the bot in this chat
I've even tried terminating the connection and restarting
Nothing worked
Till I pasted this code
nocturne mural
Yup I'm pretty sure I've tried copying two times from the link from the bot in t...
Explain that to me — what link?
oak edge
Explain that to me — what link?
-kaggle
nocturne mural
Ah, then everything’s fine. I just did a clean installation and there were no issues, so does that mean I might have this option enabled?
Since the app.py error means that the file doesn’t exist in the current directory where the commands are being executed, it’s possible that the path /kaggle/working/program_ml doesn’t have the repository properly cloned, or that %cd /kaggle/working/program_ml was never executed.
ruby valve
Can someone help me please?Whenever I try to run the Start_https shortcut and just opens for a split second and closes it’s literally like the final step too if anyone could help or lmk what can fix it.
astral perch
Is the kaggle RVC not working for other people as well?
low shard
Is the kaggle RVC not working for other people as well?
please elaborate more, tell your pc gpu, operating system, the kaggle jupyter notebook link / guide and the issue
low shard
Can someone help me please?Whenever I try to run the Start_https shortcut and ju...
start_https is from the original wokada, which isn't suggested, delete it along vb audio cable and forget youtube/video tutorials
what's your pc gpu and operating system?
astral perch
NVIDIA GTX 1650 Super. I am using the Kaggle Notebook from this guide https://rentry.co/RVC-Mainline-Kaggle
I never had any issues but now when I finish running all the cells and try to open the RVC UI, it shows that the Ngrok agent connected and everything but it couldn't connect to the local host
low shard
NVIDIA GTX 1650 Super. I am using the Kaggle Notebook from this guide https://re...
that's a pretty outdated guide, and that rvc mainline/original kaggle is at the end of life since many months, meaning it won't ever get updates again and if it works it works, if it doesn't there isn't much you can do
your gpu isn't the best for local training, especially because it has lower than 8gb vram
astral perch
Thanks, I love kaggle, it's unfortunate that it's not working anymore
low shard
Thanks, I love kaggle, it's unfortunate that it's not working anymore
the issue isn't kaggle, kaggle is just a cloud (remote good pc) computing service
astral perch
Ah I see
low shard
Ah I see
anyone can code jupyter notebooks, the issue was that the creator of the RVC Mainline kaggle jupyter notebook wasn't updated since a lot
either try the kaggle applio jupyter notebook, or try applio lightning.ai
astral perch
Do you have the link to the guides for both?
nocturne mural
.
astral perch
Thank you so much!
tawdry gullBOT
The Miles Davis Quintet - It Could Happen To You from R added to the queue (06:41) - at position 1
astral perch
I don't undderstand the Applio UI. When I paste the path of my dataset it says that it preprocessed 0 seconds of audio
simple ore
I don't undderstand the Applio UI. When I paste the path of my dataset it says t...
show a screenshot of ui
simple ore
I don't undderstand the Applio UI. When I paste the path of my dataset it says t...
in case you have trouble seeing
so like C:\training_files\
astral perch
I know, I saw
I have to paste it in the file link that kaggle provided right?
Because there is no dataset folder for me there
I tried making one but i still get the same notification
simple ore
if you're using kaggle, use dataset creator
astral perch
What's that?
astral perch
I did that but it says this. which is weird because my Audio is not 0 seconds
simple ore
then you likely upload a file in a format it does not know
".wav", ".mp3", ".flac", ".ogg"
simple ore
it is case insensitive
astral perch
what do you mean?
It works!
One more thing, the guide says that I need to make sure that the option is set to RVC V2 but I don't see it anywhere
simple ore
it is an old guide
make sure you preprocess with right settings
simple slicing, check off noise, check 'post'
astral perch
oh shit I frogot to select post lol
thanks tho
I had slicing on automatic
is that a problem
simple ore
not really, other than it is slow and uses a lot of ram
astral perch
I selected save at every 10 epochs but I can't find the folder they're getting saved to
nvm found it
thank you for your help
I appreaciated it!!!
serene dune
Is there any Real time voice changer for Iphone?
viral mason
Is there any Real time voice changer for Iphone?
oak edge
-kaggle
viral mason
it works pretty good but it sounds like it has no life in it
are you talking like a robot?
nocturne terrace
honest depot
are you talking like a robot?
im speakin normally
honest depot
are you talking like a robot?
these r my settings
viral mason
these r my settings
the pitch is probably too high most female models sound good at the max maybe 10 but 12 is ok too, and lowest 3
honest depot
the pitch is probably too high most female models sound good at the max maybe 10...
the pitch is like decent right but u can just idk how to explain it
ill show u what it sounds like acc
viral mason
and change your extra time to 2.70 to see if it sounds better
honest depot
and change your extra time to 2.70 to see if it sounds better
it just sounds like staticy and not like a real human
viral mason
could u record with snipping tool
honest depot
do these matter
honest depot
could u record with snipping tool
aight
viral mason
do these matter
idk what they do tbh so I don't touch it
honest depot
idk what they do tbh so I don't touch it
btw just to clarify
im not no asshole tryna extort money outta pimps
just tryna fuck with my friend group on an alt LOL
oak edge
i'm still gettin this @simple ore
nocturne terrace
viral mason
oh that does sound weird
honest depot
do u wanna hear what my mic sounds like
viral mason
it's probably the model itself
viral mason
do u wanna hear what my mic sounds like
sure
mic's with a lot of bg noise and if there's people talking in your bg it could mess it up
viral mason
there's a little reverb in it but that shouldn't mess up the model like how it sounded
honest depot
there's a little reverb in it but that shouldn't mess up the model like how it s...
i use this
and a blue snowball
viral mason
no clue what those are tbh but I use a mic on my valve index and it comes out fine
royal kettle
these r my settings
Have you tried turning off pitch smoothing and format shift?
honest depot
no clue what those are tbh but I use a mic on my valve index and it comes out fi...
its just filters cuz without them my mic sounds dookey
honest depot
Have you tried turning off pitch smoothing and format shift?
uhhh pitch smoothing i turned on to try and fix it
and format shift idk
ill try
viral mason
its just filters cuz without them my mic sounds dookey
that's fair
viral mason
and format shift idk
formant shift? that's like a pitch shifter
honest depot
formant shift? that's like a pitch shifter
ohh
aight ill try
also
i just realised my mic has some backround noise
royal kettle
formant shift? that's like a pitch shifter
honest depot
audacity wont pick it up brah
viral mason
oh I thought u simply mistyped formant
like what is on wokada deiteris
is it really called format and not formant
is it just typed wrong here?
royal kettle
oh I thought u simply mistyped formant
Oh you're right it was a typo- oops but yea
thick latch
Do you people actually mess with the formant? I feel like it makes things too robotic
honest depot
@royal kettle @viral mason if its just a model problem, dyk where i can get a decent korean voice
viral mason
I have no idea
I don't use human voices unless they're like Batman or the Joker
very rare occasions
visual barn
anybody know any fixes when trying to start up the batch file, it shows error code in cmd "librosa/util".
there's some posts on github yet still no solveable methods.
simple ore
anybody know any fixes when trying to start up the batch file, it shows error co...
show the message
most likely just a warning
tame oracle
is RVC still a thing?
yesh
misty temple
I'm trying to discover how the outputs were made in where the vocals are the same but the lyrics were changed.
Based on my knowledge, I can use UVR to split the vocals and instrumentals but idk what tool is being used for the lyrical change while maintaning the voicing
EDIT: Also, is UVR still a thing?
ruby valve
start_https is from the original wokada, which isn't suggested, delete it along ...
I have a Rtx 5060 and Ill delete all of it rn
tame oracle
I'm trying to discover how the outputs were made in where the vocals are the sam...
im not sure about UVR
sorry
but RVC is still a thing
misty temple
I see. No worries 🙂
tame oracle
I see. No worries 🙂
misty temple
do you have any recommendation where can I start revisiting the process again?
Idk if I should delete the repositories I have right now lol
tame oracle
do you have any recommendation where can I start revisiting the process again?
i don't sorry, i really only know about the core of w-okada
tame oracle
setup it's weird for me
Whats your setup? like which voice changer, what GPU?
knotty edge
Whats your setup? like which voice changer, what GPU?
is the model
GPU 0
Radeon RX550/550 Series
Driver version: 31.0.12029.10015
Driver date: 11/30/2022
DirectX version: 12 (FL 12.0)
Utilization 2%
Dedicated GPU memory 0.7/4.0 GB
Shared GPU memory 0.1/8.0 GB
GPU Memory 0.8/12.0 GB
alright and what voice changer are you using?
knotty edge
alright and what voice changer are you using?
w okada right?
tame oracle
w okada right?
well there is also Vonovox but Vonovox is only for NVIDIA
knotty edge
well there is also Vonovox but Vonovox is only for NVIDIA
im and amd
tame oracle
did you get it from a youtube tutorial?
tame oracle
the third one will be the best for you
knotty edge
wdym
tame oracle
just gotta get the AMD version
tame oracle
which one?
Wokada Deiteris Fork
knotty edge
Wokada Deiteris Fork
how outdated tho he was using w okada
and stuff
and it was all good
tame oracle
how outdated tho he was using w okada
because its gotten many updates since then, and this works better for AMD GPUs