#✨│ai-help

1 messages · Page 315 of 1

hallow thistle
#

And I better hope you won't make this so complicated either way.

lavish rover
#

is this a yes?

hallow thistle
#

Use WinRAR or 7-Zip to open the .zip.001 one, extract to somewhere like "D:\MMVCServerSIO".

hallow thistle
lavish rover
kind moss
#

is w okada better or worse than vcc

lavish rover
#

what would be the case in W-okada though? Does my statement hold true for the case in W-okada?

#

just curious

hallow thistle
# kind moss is w okada better or worse than vcc

Both names "VCClient" and "MMVCServerSIO" refer to the W-Okada voice changer, if you didn't know. The differences from W-Okada** fork **and the original and outdated W-Okada versions are how they implemented, W-Okada fork has more recent features while original ones are outdated.

hallow thistle
#

In W-Okada voice changer, the more "chunk" value the more delay. "Extra" value indicates how much audio quality, the more value the more audio quality, while lower one the less quality.

compact panther
#

the fact that namari is still answering this dude

hallow thistle
#

Actually, I don't use a voice changer program on a daily basis, but I know how voice changer works because so many people here keep asking for it. Yeah, that's how much I know about it.

lavish rover
hallow thistle
#

Sure, I'm more of talking about a information that hasn't been documented elsewhere, or none talked about, like how an Intel Arc integrated GPU could potentially work with W-Okada, it just doesn't have always be "how to install and use the voice changer" trope thing. Seiablob

spring chasm
#

!howtoask

patent trellisBOT
# spring chasm !howtoask
❓ How to Ask for Help
✅ Before You Ask!
  1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
  2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.
📝 How to ask?

Tell your:

  • Full GPU Name: (e.g., NVIDIA RTX 4060 8gb vram desktop)
  • Operating System: (e.g., Windows 11)
  • Detailed Description: What were you trying to do and what went wrong?
  • Tutorial Used: Link to the guide you were following.
  • Screenshot: A picture of the full error message.
🚫 Prohibited Topics (We Will NOT Help With These)

To maintain a legal, safe & ethical community, we will NOT provide help for:

  • ANY illegal activities.
  • NSFW/Porn.
    Requests for these topics may be ignored, not helped and result in moderation action.
<:matsuripray:1159685390156967936> Community Expectations
  • Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
  • English Only: Please keep all conversations in English.
  • Don't Ask To Ask.
lavish rover
#

I'm running Vonovox with a RTX5090 on Windows 11. I've followed the guide on AIhub to set it up. I'm satisfied with the quality. I just have two questions:

  1. I notice it's only just a small difference in delay even if I crank the block size to 1.0(max). Does keeping the block size at max means the best sound quality and accuracy? (I tested the lowest block size value that does not sound choppy, I genuinely cannot tell the difference in quality.)

  2. the output volume is a bit low, should I turn up the input volume or output volume to avoid introducing noise/ artifacts?

void flume
#

No, it will only reduce quality after the block size gets too small. There is no increase in quality after a certain amount.
Better Quality and Accuracy is what the Extra regulates.

#

yeah actually

lavish rover
#

I have 'extra' set to max.

lavish rover
hallow thistle
#

I use laptop's integrated microphone. mikascared

lavish rover
void flume
#

It's not that big, a dynamic mic I mean, but it does have a size lol.
but the issue is cross communication; you want to avoid having your friends be heard by your microphone

patent trellisBOT
void flume
#

So its mainly a headphone issue, or if you really want to use speakers, volume and noise gate- but I guess in that case you're going to have a terrible time with settings if you use a laptop, since everything is very close to eachother.

lavish rover
void flume
#

cross talking basically.

#

the microphone catches all audio playing, and if it is a voice, then the voice changer will run it

lavish rover
#

cross talking?

hallow thistle
#

I'm not sure how to interpret that. If you mean by "your microphone is too close to your speakers", I kind of understandable.

lavish rover
#

I feel like my brain is not braining

hallow thistle
#

Got anything you wonder how some specific features in Vonovox work? There's a Discord server about "Vonovox" especially.

lavish rover
hallow thistle
lavish rover
lapis kiln
#

hell

#

o

#

What is the benefit of this?

Screenshot_2026-01-22_150106.webp
#

I put it on

hallow thistle
# lapis kiln What is the benefit of this?

This setting "forces the voice changer to use FP32 the 32-bit floating point with your PC GPU". This will make the GPU to process more, sometimes will give better audio quality but often more audio delay compared to non-FP32 mode.

lapis kiln
#

Why does the volume rise when I speak? perf

Screenshot_2026-01-26_180834.png
#

It can also reach 135

#

How do I make it stable in the 60 - 90

#

??

hallow thistle
#

Um.

#

That's what you'd expect when you turn "force FP32" on. To put perf number at 60 - 90 you would have to disable "force FP32", when disabled there will be less audio delay but potentially lower audio quality at the same time. Try set chunk number to around 180 ms.

lapis kiln
#

If you advise me to put off FP32

hallow thistle
#

There's a trade off, by the way. Ado

oblique briar
#

CAN U HELP

#

how do i fix delay

lapis kiln
oblique briar
#

anyone can help?

sullen heron
void flume
# oblique briar

Try turning your speakers off or disconnecting them and see if it still happens. You can use audacity to record an audio file, and listen to it after reconnecting your speakers.

oblique briar
void flume
# oblique briar how/

never mind I guess.
Is it just Realtek HD Audio that plays the audio twice? Did you try recording from the Cable without listening to it?

oblique briar
#

no i listened

#

it still bad

#

but on voice mod its good

void flume
oblique briar
#

is it better?

void flume
#

It's faster and more stable

oblique briar
#

ok

#

im download rnm

void flume
#

You only get 1 cable with the lite version, but it's enough. Open up VAC control panel as admin, and change the settings to this:

#

and then click on Set in the bottom right

oblique briar
#

how i oepn it

void flume
oblique briar
#

i

#

extracted

#

didnt open

#

up

void flume
#

it looks open

oblique briar
#

wont let me edit

#

to urs

void flume
#

it needs to run as admin to do that, but.. I guess its fine

#

most of the settings are good enough

oblique briar
#

this will fix

#

the cutting out and lag?

void flume
#

I don't know.

#

So now you have a Line 1 in your audio drivers

#

Try setting the Output of the voice changer to Line 1
and the input on Discord to Line 1

#

Also double click this in Sound settings, and make sure it is set to 48000 sampling on the third tab

image.png
#

It should be 2 channel, 24 bit, 48000 sampling mode

#

same as in vac control panel basically, and same in the voice changer

void flume
# oblique briar the cutting out and lag?

There is always some lag, as people mentioned.
like, about a second... slightly over half a second at least. (700ms or something)
you can't fix the lag entirely, because its going through complicated AI tools before the audio is put out.

oblique briar
void flume
#

but the double audio should be gone at least

void flume
#

input is where your microphone goes

oblique briar
#

what about thisd

quick summit
#

Am i the only one having issues with Applio?

void flume
# oblique briar

Depending on what you pick, it could cause the echo / repeat of the audio.

oblique briar
void flume
oblique briar
#

this normal?

void flume
#

speaker is where you put your earphone

void flume
#

Speaker makes sound, microphone receives sound.

oblique briar
#

ye

#

good?

void flume
#

Looks good

#

Is there still odd audio output, like hearing yourself twice?

#

... I guess no? okay case closed hopefully

oblique briar
#

and show u

#

i just dont sound like a girl idk why

viral mason
#

Did you click start

oblique briar
#

hello

void flume
#

The only thing I can imagine here at this stage is that it is the microphone picking up sound from the headset. (echo)

#

as to why it is softer than with VB-Audio, idk- could be filters or something. VAC doesn't strengthen the audio at least

#

You may need to mess with the volume and use a noise gate, but it may not be enough, depending on how sensitive the mic is

#

The first sound should be the loudest, but it for some reason is the softest, so there is a gain filter powering up the audio somewhere in the loop

oblique briar
#

okay

oblique briar
void flume
#

I want to eat dinner

oblique briar
#

i think my voice is to deepin

#

for girl fliters'

#

:c

void flume
#

Just ... start by disabling any Gain. (input volume 100%, output volume 100%, gain 0), then see when the echo happens when you raise the volume of your headphone.

#

echo meaning, that the voice changer is producing the output more than once

#

Then you need to find a sweet spot of where you can hear things, but for your microphone it is too soft to hear

#

if you cannot find this, then you need to use Noise Gate to filter out softer sounds.

void flume
# oblique briar for girl fliters'

Pitch needs to be 12-ish, and you need to do the rest yourself, talk on a higher pitch, etc.
you can also raise it above 12 if it is difficult, but if you have little control over your voice it will sound weird.

#

pitch to 24 if you want to be lazy, but then you need to talk with a deep voice constantly

#

You can offset it a bit, but then your C note will sound like an E note or something, you basically want to keep the notes correct while offsetting the octaves

oblique briar
#

oksy bvet

craggy magnet
#

guys what do i do

image.png
#

i tried switching browsers

#

i deleted settings

#

idk

viral mason
#

If you deleted anything just reinstall

craggy magnet
#

Pk

#

Ok

oblique briar
#

@void flume yeah

#

it doesnt sound believeable enough tbh

craggy magnet
#

i reinstall and now when i launch the mmvcs it pops up for sec and close down

light swift
#

i need do something here?

image.png
#

for input i need vr cable or my micro

void flume
#

output cable

dusty rune
#

guys where do you put your d g pths in applio?

viral mason
dusty rune
dusty sparrow
#

Does anyone know how to fix the TTS in Applio? It does not work in collab nor locally. Also seems like some other tools online that use the Microsoft voices ( ? ) are maybe bugged too.

light swift
#

although others can't hear me with such a voice

viral mason
light swift
#

There is another problem: I have already tested 3 voice models and they practically do not change my voice at all, only an echo appears from behind the speakers.

viral mason
#

what is ur input and output

light swift
viral mason
#

oh

#

I can't read them

#

your input should be your regular microphone, and output should be your virtual audio cable

#

vac lite

light swift
#

Okay, but what should I do if my voice doesn't change at all?

viral mason
#

I'm not sure

light swift
#

I thought that this was the only program where you could practically not hear the AI, but no matter what voices I use, they change literally by half a tone

viral mason
#

I've never had that issue so I would ask a helper such as namari

light swift
void flume
#

the virtual cable is used to transfer audio from one place to another. You need it or something like it because Windows doesn't support such things by default.

#

The voice doesn't change? That can happen for a number of reasons.

#

Try restarting the server. There are bugs that happen on first launch.

#

Sometimes, restarting the entire system (in case drivers changed) may be needed.

light swift
void flume
#

Other than that, you should check whether or not the webbrowser page loading the voice changer has permission to access the microphone,
(the mic's signal may be outputted by the speaker through another piece of software, not just the voice changer; you may need to investigate whether this is happening or not.)
Make sure passthrough disabled.
And of course whether or not the voice changer is running.

light swift
#

this change

#

but little bit

#

very little

void flume
#

Set the pitch to 12 (if you're a guy) for a girl voice to work
also the output is going to be softer by default, so you may want to increase the output volume.

#

that girl voice model speaks as if it is whispering, from what I understand, so- it's going to sound soft in my guess

viral mason
#

literally throwing up from the meantion of that nvm u fixed it

light swift
void flume
#

it says 404 (page doesn't exist)

viral mason
light swift
#

Okay, just one question: why is the community -11 in Virus Total?

image.png
viral mason
#

what does that mean?

light swift
#

idk like -rep

void flume
#

I don't get those same results...

void flume
light swift
#

maybe

light swift
void flume
#

oh wait you uploaded the zip file

void flume
viral mason
light swift
#

same program other link

void flume
#

I'm glancing at VirusTotal currently. All of those -11 rep points were placed by the same user.
also that user is pretty.... odd

light swift
#

hmm okay

void flume
#

His server is down, so I don't know why he came up with those conclusions

light swift
#

any guide for vac470lite?

void flume
#

It's plug and play basically, you just need to install it

#

The scoring is pretty aggressive.

light swift
viral mason
#

setup64

void flume
#

I looked at the reports for the bad score, it looks like the analyst had an infected ZIP file with code that is way different, causing different behavior

#

I mean, we can all look at the source code of batch files, and these don't connect to servers in the US.

void flume
#

Don't pirate basically.

light swift
#

okay i download

#

what now?

viral mason
#

run this file

image.png
light swift
#

but i download already

viral mason
#

oh lol

light swift
#

i mean what to do on discord

#

for people hear me

viral mason
light swift
#

If I put the virtual cable on input, they don't hear me. If I put it on output, I don't hear. That's weird.

#

or for this to work I need to enable the ai model first?

light swift
viral mason
light swift
#

like nothing

image.png
quasi condor
light swift
#

i don`t wanna

#

i wanna hear other and for other hear me on discord

quasi condor
#

im sorry what? 😭

#

@viral mason is this guy okay

light swift
#

bad English maybe

#

I want others to hear me on Discord, but when I connect to a virtual cable, nothing happens.

#

so I don't understand

compact panther
# light swift I want others to hear me on Discord, but when I connect to a virtual cable, noth...

Think of a Virtual Audio Cable as a USB slot.

On its own, it won't do anything. You have to "plug in" your headset, or microphone or whatever, into that VAC, or else it won't transmit anything.

If you use a Voice Changer, in there, you must set your microphone as input, and Virtual Audio Cable as output. Then, in every other app, you select Virtual Audio Cable as input, wherever you want to use the Voice Changer.

#

Example:

In Voice Changer

Input: <your mic>
Output: Virtual Audio Cable

In Discord (or any other app)

Input: Virtual Audio Cable
(Output: <your headset>)

real beacon
neon mulch
#

Anyone know why my i have vb cable and my mic isnt picking up any sound

#

the vb cable is set to input

compact panther
compact panther
#

btw, do not ever set input AND output to Virtual Audio Cable within the same app

hallow thistle
#

Setting audio input to "Line 1 (Virtual Audio Cable)" and then output to "Line 1 (Virtual Audio Cable)" is an incorrect one, regardless of which program it using. akanesmile

patent locust
#

guys how do I solve such problem?

#

the output voice become chaotic

light swift
#

already

#

but still not working

hallow thistle
# patent locust

This "RVC WebUI" likely being an older version. Try Applio RVC the more recent "RVC WebUI fork" instead.

#

-rvc

patent trellisBOT
sullen heron
#

Is a batch size per gpu of 6 good for a 4060 8gb? im training a model at 200 epochs with around 17-20 minutes of data

hardy yew
#

seems okay, although with that much data i'd probably choose 4. Don't have too much experience yet, but I've seen people recommend higher batch sizes (like 8) when you have bigger datasets (e.g. 1h and more)

#

the training should be fairly quick anyway, so i suppose it doesn't hurt to try both, really

sullen heron
#

i tested it at 40 epochs at a batch size of 8 and it sounded really really accurate to the voice im trying to clone

#

but i saw people say to use 6 instead or 4 for a gpu like this, so it got me overthinking if putting it to 8 would ruin the quality

quasi condor
#

batch size 8 is goood

hardy yew
quasi condor
analog obsidian
#

i'd not recommend high batch sizes to someone who just started to make models

sullen heron
#

Im monitoring my vram though people say to turn it down when it leaks to shared memory and its leaking into shared memory right now even at 6

quasi condor
hardy yew
quasi condor
analog obsidian
#

17-20 minutes is not a large dataset

quasi condor
#

my mj model was 1 hour tho

#

so large?

sullen heron
analog obsidian
quasi condor
#

4 is good

#

now i go to bed night peeps

sullen heron
# analog obsidian use 4

Will i have to retrain my model all over again? Or will pausing then changing the batch size be fine?

hardy yew
#

don't change batch size mid-training

analog obsidian
sullen heron
#

yuh oh

hardy yew
#

oh, Lyery has much more experience so definitely listen to their opinion more than mine

#

i've read before that this was disadvised but perhaps i'm mistaken

analog obsidian
#

🤦‍♂️

#

there are some misinformation about the batch size too

sullen heron
#

thanks a lot for the help guys ❤️

hardy yew
#

i wish that kind of knowledge was documented somewhere, over past week or two i've scrolled through a lot of info and it's difficult to separate the good parts from the bad

analog obsidian
#

the problem with the high batch sizes is that they overshoot, but in rvc i havent noticed any problems with them

#

but for finetuning i'd not use anything above 32

#

when i tried batch 128 in a pretrain, it made it explode
if your model is not exploding, ur good lol

analog obsidian
#

if you notice the model sounds good and everything it's fine, means the batch you chose is perfect

#

no need to overthink

hardy yew
# analog obsidian back then people were saying the low points of the tensorboard graphs were mode ...

I know that looking just at the loss graphs isn't enough to choose the best checkpoint, but I would expect the charts to act as guidelines in which areas to look for the checkpoints. Like, it can be noticed when the initial training stage with huge gradients ends and where's the local minimums which could suggest some decent checkpoints. And later on the training could be monitored for regression. Is that approach wrong? I'm trying to get into some good habits

sullen heron
#

One last question, how may i fix the very robotic and metallic “S’s” in models?

hardy yew
#

Would very much appreciate any tips on how to interpret the tensorboard to make good use of it during/after training

analog obsidian
sullen heron
analog obsidian
#

i only check the tensorboard logs when im training a pretrain from scratch and i need to know if my model is exploding

#

which doesnt happen with normal rvc training, unless you do it on purpose

hardy yew
#

So just train for some decent amount of steps while saving checkpoints every X epochs and test & compare the checkpoints afterwards for choosing the best results?

analog obsidian
sullen heron
analog obsidian
hardy yew
#

very valuable info, thanks!

analog obsidian
patent trellisBOT
sullen heron
analog obsidian
#

i noticed batch size 4 has decent breaths and esses, what if you try that?

sullen heron
analog obsidian
#

the step number is listed after the epoch number
for example: model_10e_3000

sullen heron
#

Gotcha, will keep that in mind

analog obsidian
#

click fresh training and change the batch to 4

sullen heron
#

The voice im trying to replicate is an ASMR youtuber who does a lottt of whispering and some very light talking so im trying to find settings that would complement well to the many emphasized “S’s” they do and their breathing, as well as the very light talking

analog obsidian
# hardy yew very valuable info, thanks!

yes, something I always recommend is not to blindly believe what people say about rvc, it's best to experiment and find the best parameters for your model on your own xD

sullen heron
analog obsidian
sullen heron
analog obsidian
hardy yew
#

getting best of both worlds

analog obsidian
#

and send me a screenshot of the webui because i dont have mainline installed lol

sullen heron
analog obsidian
#

the applio code is the same exact of the rvc webui

#

so both produce the same result

#

yea i recommend it

sullen heron
analog obsidian
#

don't run the preprocessing again

#

be sure it's v2 and 32k

#

then decrease batch size to 4 and click train model

sullen heron
#

Do i change the pretrain? Since I am using LegacyCore 1.5

analog obsidian
#

you can also enable this so it doesnt save every G and D files

image.png
#

and enable this

image.png
analog obsidian
sullen heron
#

Alright, do I try switching if the final result doesn't meet what I want?

analog obsidian
#

sure, actually thats what i always tell to ppl, first use the default pretrain, then experiment with others if you didn't like the first result

sullen heron
#

What would you recommend for the save frequency?

#

I use 10

analog obsidian
sullen heron
#

I am using tensorboard so would i also need to delete anything to reset the graphs or do i just train it

#

sorry if theres too much questions by the way ill stop if you want 😅

analog obsidian
#

yes, delete anything inside the eval folder

sullen heron
analog obsidian
sullen heron
#

The files in eval?

analog obsidian
sullen heron
analog obsidian
#

delete these

sullen heron
#

I did delete the files in eval before starting the training earlier though

analog obsidian
#

also delete G and D files, start again

sullen heron
#

How can i test specific checkpoints without ruining any training progress?

#

do i just X out of the terminal, test the checkpoint, load up rvc webui again then continue training?

rugged wren
#

Hey 2 things

  1. How do I make it so when I speak on the actual voice changer app I can hear myself while doing it? It happened and then it just stopped randomly
  2. My gpu isn’t getting recognised, my gpu just says cpu and that’s it, I have a rx 9070xt
halcyon palm
#

скиньте 500 рублей пж

sullen heron
#

This is what I have at 11k+ steps, sustained S's sound really bad

#

Could this be caused by a low quality dataset or does it just need more training?

analog obsidian
#

and nsf hifigan

#

i think you can add more data to improve it slightly but it's still gonna be unrealistic

sullen heron
analog obsidian
#

its not perfect

sullen heron
#

Ah i understand it now, thanks

#

Any way to improve it or is adding more data my safest and best bet?

analog obsidian
#

yep, but keep in mind it will never be 100% perfect and realistic

sullen heron
#

Alright, thanks for being a big help for my first model cat_cool

rugged wren
tame oracle
patent trellisBOT
# tame oracle !howtoask
❓ How to Ask for Help
✅ Before You Ask!
  1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
  2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.
📝 How to ask?

Tell your:

  • Full GPU Name: (e.g., NVIDIA RTX 4060 8gb vram desktop)
  • Operating System: (e.g., Windows 11)
  • Detailed Description: What were you trying to do and what went wrong?
  • Tutorial Used: Link to the guide you were following.
  • Screenshot: A picture of the full error message.
🚫 Prohibited Topics (We Will NOT Help With These)

To maintain a legal, safe & ethical community, we will NOT provide help for:

  • ANY illegal activities.
  • NSFW/Porn.
    Requests for these topics may be ignored, not helped and result in moderation action.
<:matsuripray:1159685390156967936> Community Expectations
  • Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
  • English Only: Please keep all conversations in English.
  • Don't Ask To Ask.
spring orchid
#

why i cant add effect to my mic input on wokada client mode ? on server works fine but on client is not picking up my mic with my effects

viral mason
#

client has been broken since november 2025

#

no explanation just stopped working for everyone

#

only server mode works

void flume
#
  1. use the directml release (dml) of the voice changer, not the cpu release.
random plover
#

Is it normal that when i hear myself, i hear it delayed? and does it affect conversations like when im talking to someone is it delayed too?

ember lily
paper robin
#

how does one train an ai voice? like acn i use an mp3 to make it sound like someone?

viral mason
patent trellisBOT
viral mason
#

read up on this guide, u would need to use applio to train the voice and also would need to clean the audio so there's no music or anything

paper robin
#

thank you!

viral mason
#

you're welcome! u can use UVR to isolate vocals from audio as well as sites like mvsep or minus x

hallow thistle
rugged wren
#

It never ends really, a echo that never ends

#

I fixed it, also I use do I use 0: “AMD Radeon RX 9070 XT (DirectML) or do I use “1: AMD Radeon(TM) Graphics (DirecetML)

void flume
#

RX 9070 XT (DirectML), the other one is your iGPU. You could use both, but the graphics card is likely faster.

void flume
# rugged wren It never ends really, a echo that never ends

Earlier I thought you wanted the ability to listen to yourself, which is why I send what I did before.
Echo issues tend to be on the hardware side, usually at least. Too many people don't realize what their speaker broadcasts can be heard by their microphone, and usually will be. That combined with audio enchancement causes a loop where on each repeat the audio gets louder.

rugged wren
rugged wren
rugged wren
hallow thistle
patent locust
#

😲

hallow thistle
#

"Applio RVC" is the most recent RVC UI that most people here use. What do you mean Applio RVC won't gonna fix robotic audio either way?

patent locust
hallow thistle
patent locust
#

ok

#

why it only can be installed on c drive

#

my c drive is full

hallow thistle
#

Try extract "Applio RVC" to somewhere that's not always be drive C.

image.png
patent locust
#

thanks

#

you are awesome

hallow thistle
#

Applio RVC on Kaggle, aside from using website "loca.lt" to access file directory of that Kaggle environment for dataset audio files, I also use it to upload "multiple voice model files" there too.

image.png
dry coyote
#

is vonovox necessarily faster than deiteris's w-okada fork?

hallow thistle
viral mason
dry coyote
#

like what more does vonovox have to offer that wokada doesnt have

hallow thistle
dry coyote
hallow thistle
#

Unsure and inconclusive. misc_shrug

dry coyote
#

i use w-okada fork right now, and i do notice "noise" that the voice changer gives

#

so i usually have to use a low pass fulter to get rid of that but it slightly destroys the resonance

#

im asking if vonovox just increases the audio sample rate (kbps) or if it increases the "realisticness" of the RVC

hallow thistle
#

Actually, I can focus on settings and things,** not always cared too much about audio quality **that either program gives, so while I've never tested Vonovox and W-Okada fork at my own, I can answer that way. Moodeng

hexed wolf
#

so far imo vonovox sounds a good bit better than the okada forks

#

using the beta version of vonovox btw

hallow thistle
#

You know, Vonovox and W-Okada** don't **use bitrate (Kbps) like what CBR (constant bitrate) audio files do unless record as a file otherwise, but rather happen with a voice model quality and how you set block size, extra time and crossfade on Vonovox.

hexed wolf
#

it also allows for configuration of the voice texture going into the ai model, so if you have a model that's supposed to sound very gritty and deep ( Master Chief for example ) you can make it sound much closer using that setting.

#

haven't tested going the other direction with voice texture yet but i would think it would help with smooth voices going the other way.

patent locust
#

guys how to install voice model on applio RVC

#

where should I put the model file

hallow thistle
#

I just closed Applio RVC. midorisob

#

On Applio RVC, in "Download" menu, you'll see some sections to upload your voice model(s). https://cdn.discordapp.com/attachments/1159290139609137264/1458081997040783452/image.png?ex=697ab06d&is=69795eed&hm=5dfd7085469aac4f312d628fbf054be21a953ee8b82c4cda924bffa87e541f4a& Then go back to "Inference" tab, click a big "refresh" button. https://cdn.discordapp.com/attachments/1159290139609137264/1455545707204182016/image.png?ex=697ab0d3&is=69795f53&hm=ccb7944101089ffde5b555dd90bb1387e4a9e3f56f0295d0973d5c76ed585d69&

patent locust
patent locust
#

😭

hallow thistle
patent locust
#

can W-Okada or Vonovox also do AI voice changing based on a input audio file?

hallow thistle
#

If that one voice model came from 2023, it's likely the case, or perhaps the model was trained with low-quality dataset audio and settings from start. Realtime voice changers mostly use your microphone live, a few voice changer versions can process from a file like non-realtime Applio RVC, but some voice changers usually lack that feature and that's about it.

patent locust
#

ok thanks

dusty zodiac
#

It doesn't work now ?

#

facefusion.py run: error: argument --execution-providers: invalid choice: 'cuda' (choose from 'cpu')

winged star
#

-rt

patent trellisBOT
# winged star -rt
💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

random plover
#

Can someone help me please

#

Nothing works for me

winged star
#

-rt

patent trellisBOT
# winged star -rt
💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

hallow thistle
cloud igloo
#

my okada always has to be open when im using it. is there any way to use it in the background

hallow thistle
patent trellisBOT
# hallow thistle !howtoask
❓ How to Ask for Help
✅ Before You Ask!
  1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
  2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.
📝 How to ask?

Tell your:

  • Full GPU Name: (e.g., NVIDIA RTX 4060 8gb vram desktop)
  • Operating System: (e.g., Windows 11)
  • Detailed Description: What were you trying to do and what went wrong?
  • Tutorial Used: Link to the guide you were following.
  • Screenshot: A picture of the full error message.
🚫 Prohibited Topics (We Will NOT Help With These)

To maintain a legal, safe & ethical community, we will NOT provide help for:

  • ANY illegal activities.
  • NSFW/Porn.
    Requests for these topics may be ignored, not helped and result in moderation action.
<:matsuripray:1159685390156967936> Community Expectations
  • Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
  • English Only: Please keep all conversations in English.
  • Don't Ask To Ask.
hallow thistle
#

Here's an example on how to detail your issue:

W-Okada voice changer doesn't work on my PC
PC GPU: NVIDIA GeForce RTX 3080
Operating system: Microsoft Windows 11
The reason: W-Okada doesn't work, I tried to follow the tutorial but still doesn't work.
Tutorial I used: a YouTube video link you followed.

sullen heron
#

hi! how do i make whispering work well on vonovox?

#

it doesnt have an rmvpe_onnx option for f0 method like wokada does, and thats the one that usually makes the whispering sound real

hallow thistle
#

There are big trade-offs between W-Okada voice changer and Vonovox. aronathink

sullen heron
#

ooh

sullen heron
hallow thistle
#

Vonovox likely has regular rmvpe model because it's supposed to work with NVIDIA GPU. For "rmvpe_onnx" F0 model option, check out Tg Develop's W-Okada fork.

winged star
#

@hallow thistle Hello

#

Can you help me to train a voice ?

#

I put my video on it

#

I got now the track

hallow thistle
winged star
#

Oh ok

#

Can I send you the .mp4

#

and you tell me if this one

#

is good ?

rugged wren
winged star
#

I'm trying to train this voice

hallow thistle
# winged star Can I send you the .mp4

Well, actually I'm not a model maker member, so for training a voice model it's better to ask a fellow model maker here instead. And make sure you** don't** send any copyrighted material (especially audio) to this server, regardless of what that supposed to be.

winged star
#

Yes

hallow thistle
winged star
#

I will ask to a model maker

#

ty *

#

I'm trying first

#

the training part

#

So you don't really know what configuration I'm suppose to do

#

On the training part ?

hallow thistle
#

I literally just implied like that. If I say "I'm not a model maker" it means I don't know how to train a model. You good?

winged star
#

yes im good and you

#

?

hallow thistle
hallow thistle
winged star
#

look the documentation for me

hallow thistle
winged star
#

you want me to add you in friend ?

hallow thistle
# winged star you want me to add you in friend ?

The what you mean by "adding friend"? I didn't say I want to help you in direct message. You could wait for another model maker to answer. And while I can provide some basic settings for "inference" part in Applio RVC, it doesn't have always be me to answer anything. Do you understand or still wanting answer?

rugged wren
#

I mean it looks ok, it’s the same as my other one it just stopped working from last night, is it working fine…

hallow thistle
# rugged wren

Oh, that's not Tg Develop's W-Okada voice changer fork b2397, it's Deiteris' W-Okada fork b2332.

rugged wren
#

What one am I meant to download… it’s so confusing

#

I had another one before it didn’t work at all but I have a feeling that was the right now

hallow thistle
rugged wren
#

Oh yeah I had that one

#

It didn’t work at all

#

So I deleted it, thought I had the wrong one and I couldn’t send images

rugged wren
hallow thistle
#

Some flaws I noticed from your screenshot are:

  • Your output/monitor audio devices are confusing, the "monitor" one isn't your speaker either which is why you won't hear the program through a supposed speaker.
  • Chunk and extra values are of default settings, which** should** be set to some value else.
rugged wren
#

So I want this one?

rugged wren
#

If it’s on anything else I can’t hear anything so I guess that’s ok

#

But that isn’t the problem if I record my audio or whatever and listen to it, it’s just silent but yeah I’ll get the other thing

#

what one do i even get? the bottom?

Screenshot_2026-01-28_140536.png
hallow thistle
#

voice-changer-windows-amd64-dml.zip

#

The initialism "dml" refers to DirectML, the one that will work with AMD Radeon RX GPU. The "CUDA" one is for NVIDIA GPU only, while "CPU" means CPU-only no-GPU.

winged star
#

I'm looking for the audio analyzer

rugged wren
winged star
#

an idiot

#

u are just learning

#

@hallow thistle right ?

rugged wren
#

Beats me

winged star
rugged wren
#

i dont get this one but its not working for me

Screenshot_2026-01-28_141332.png
rugged wren
winged star
rugged wren
#

I’m so confused

hallow thistle
hallow thistle
# winged star

Because your Applio RVC UI language is set to "French", meanwhile almost everyone here uses English, so many settings might look confusing to you.

rugged wren
#

can you send a link please if its ok?

hallow thistle
winged star
#

@hallow thistle

image.png
#

The sample rate its this one

#

There is no 22050

rugged wren
#

It seems to be working now! Thank you

hallow thistle
rugged wren
#

Bro this is fucking amazing!

hallow thistle
# winged star There is no 22050

RVC never had an option to train a voice model in 44100 (the double of 22050). There are 32000, 40000 and 48000 the common across many RVC voice models, with 32000 being the lowest one possible. For help about "how to prepare dataset audio" and "how to train a voice model", just letting you know** for second time** already, there are some members here with a pink "model maker" role in their accounts.

hallow thistle
jovial dock
#

what are ther most recomended settigns for w okada fork

#

or best

limpid jackal
#

hey guys i havent made an ai cover for years.. where can i make one?

hallow thistle
hallow thistle
limpid jackal
rugged wren
jovial dock
hallow thistle
# jovial dock what about advanced settings

"Crossfade length" at "0.15 s" while turn "force fp32" on. Not always recommended settings, but these settings can help make better audio quality in voice changer, with a trade-off of a bit audio delay.

jovial dock
#

thans

limpid jackal
hallow thistle
limpid jackal
hallow thistle
#

The "Applio RVC' in question looks like this as an app.

image.png
hallow thistle
rugged wren
#

It gets like a 10 second delay from when I speak to it repeating

hallow thistle
rugged wren
#

Oh I’m so stupid thank you

limpid jackal
hallow thistle
chilly furnace
#

tho not including free tier limits + most spaces arent updated anymore

#

dead end

#

'less you set up a notebook

viral hill
#

What’s the best AI tool right now to change a singer’s voice in a song?

viral mason
#

-rvc

patent trellisBOT
viral mason
#

Here's the docs to look over, applio can be used both locally installed as an app on your PC and also on a browser on something like Google colab or Kaggle (kaggle is better)

long canyon
#

says theres an issue with voice conversion and i need to check command line window for more details. not sure how to get to the command line

gusty moth
#

-colab

patent trellisBOT
# gusty moth -colab
📒 Google Colab Notebooks

Google Colab is a Cloud (Remote Good PC) Service. While the Free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

• **Music Source Separation Training (Inference)**

by Jarredou & Makidanye
Google Colab

tame oracle
#

!howtoask

patent trellisBOT
# tame oracle !howtoask
❓ How to Ask for Help
✅ Before You Ask!
  1. Check Docs & Guides: Your answer may already be in the AI Hub Docs or the https://discord.com/channels/1159260121998827560/1159513888199540817 channel.
  2. Search the https://discord.com/channels/1159260121998827560/1192011222023950368 : Look for existing posts that solve your issue. Do not invade someone else's post.
📝 How to ask?

Tell your:

  • Full GPU Name: (e.g., NVIDIA RTX 4060 8gb vram desktop)
  • Operating System: (e.g., Windows 11)
  • Detailed Description: What were you trying to do and what went wrong?
  • Tutorial Used: Link to the guide you were following.
  • Screenshot: A picture of the full error message.
🚫 Prohibited Topics (We Will NOT Help With These)

To maintain a legal, safe & ethical community, we will NOT provide help for:

  • ANY illegal activities.
  • NSFW/Porn.
    Requests for these topics may be ignored, not helped and result in moderation action.
<:matsuripray:1159685390156967936> Community Expectations
  • Be Polite & Patient: Our helpers are volunteers. You may ping the Helpers role once.
  • English Only: Please keep all conversations in English.
  • Don't Ask To Ask.
pallid cedar
#

-huggingface

patent trellisBOT
mighty lynx
#

i have a problem

image.png
viral mason
mighty lynx
viral mason
#

Idk what that even is that you're using

mighty lynx
#

can't see you the problem is "traceback (most recent call last)"

viral mason
#

That looks old as hell man

#

Where'd u get it

analog obsidian
#

🦈

mighty lynx
#

OH

viral mason
#

Yikes

mighty lynx
#

why didn't you tell me then??

viral mason
#

Ngl I didn't even know what I was looking at I'm sorry

mighty lynx
#

ik, its okay
because i downloaded newest rvc gui and it says python was not found

#

but i have python

#

thats why i used that old ass 2023

viral mason
#

What are you wanting to do with it anyway?

#

If you're looking to do covers I'd just use applio

mighty lynx
#

well, i kinda looking to do covers

median knoll
#

Hi everyone, I'm using the RVC WebUI that you can access from computer files, it's been a while since I used it, but I just tried to import a voice model and it's not appearing in my inferencing voice dropdown even though I put the INDEX and PTH files where they should be? is the UI broken?

viral mason
#

If you're wanting to do covers just switch over to applio as it's the most up to date for that kind of stuff

#

-rvc

patent trellisBOT
median knoll
viral mason
#

I haven't seen that in years

#

That's so old..

#

It really is recommended tho to.use something up to date tho over something simple

median knoll
#

It can't be that old I used it 4 months ago 💀

#

But I'll keep that in mind thank you

viral mason
median knoll
viral mason
#

It looks somewhat familiar as I used to use something like that a couple years back but at the same time looks unfamiliar

#

Layout is quite similar to applio just without the Chinese text

median knoll
viral mason
#

Pretty much, it has a little bit of extra features as it's getting updates somewhat often

#

You can also run it locally or on browser which is cool

median knoll
#

I thought locally was easier because I didn't think the system could break, but maybe mine did

viral mason
#

I don't personally have the download for the local version but you could ask someone like Namari as they may know more

viral mason
#

It might be running on an older one and you have a newer one or other way

compact panther
#

hey y'all i have a different kind of AI/bot question... coming here from #🧬│ai-chat

median knoll
compact panther
#

can someone give me a good pointer or guide on how to write an effective game player bot that acts based on visual input (=screen)

#

and interacts with the game via emulated key inputs

#

i mentioned this in the other channel too: this is an entirely offline project, so no, i'm not cheating

#

(i'm even TASing this game)

viral mason
#

There's an app called replay but you need a weights.gg account to use it and they're shutting down

median knoll
viral mason
#

Ohhh

#

<@&1159293140440723499> cough cough

compact panther
#

really? nobody here sharing my interest?

viral mason
#

I swear nobody reads the mo advertising rule

compact panther
#

the only other sources i've got are google and chatgpt

viral mason
#

There's some AI thing that hasn't released yet but it's supposed to play the game for you or something, that's all I know of it

compact panther
#

well okay

#

google and chatgpt it is skull_sob

#

thanks-ish

viral mason
#

It's from razer, project ava

#

🤷‍♀️

#

Anime character in a tube

#

Strange to me

compact panther
#

so it's basically

#

chatgpt in a jar

#

with fancy lighting

#

oh i'm sorry, grok in a jar with fancy lighting

#

||shut up sapphire||

viral mason
#

They could have chosen anything but grok 😭

#

Grok is cool but really isn't in the best place rn with those images it has been making

wise wren
#

Maybe this is me being dumb but I'm unable to open them with 7z

sleek comet
#

Hi, i have a question. It's possible use voice changers in mobile? If it's possible, can please help me how?

#

(I don't know if this channel it's correct to ask)

void flume
void flume
wise wren
void flume
#

You pinged someone specific, but I wasn't sure if you wanted strictly them to reply or not

#

but its taking a while seemingly before they react

wise wren
#

oh also

#

Im having another issue, im using a 5070 ti and when I try switching from my CPU to it I get a "An error occurred during voice conversion. Check command line window for more details." popup

#

Would you happen to know what could be causing it?

void flume
#

Not really, it could be a bug on the specific fork of w-okada voice changer you're using.

#

For the 5070 ti, I think vonovox is a better option as voice changer

#

but on w-okada, most people use tg-develop's fork. which does have rtx 5000 series support.

wise wren
#

I used the one that user recommended for the 5070 ti which I replied

wise wren
void flume
#

There are pros and cons with vonovox, as well as tg-develop's fork of w-okada- it's basically a matter of personal taste

wise wren
#

thank you!

void flume
#

but vonovox currently has less bugs and might be faster than w-okada for you

wise wren
void flume
#

(not sure since i don't have an rtx to test it with lol)

wise wren
void flume
#

Good luck anyway ;p

sullen heron
#

Does anyone know how to fix occasional raspiness in tg develop's w okada fork?

#

Sometimes, when I talk normally it produces a super duper raspy result

#

Even though there isnt anything wrong with my speech or mic

sullen heron
sullen heron
#

I'm on a 4060 by the way,, I would choose vonovox but tg develop's fork has rmvpe onnx f0 method which is what i need

void flume
#

It could be the voice model.
you can try to optimize towards quality (extra 2.7, crossfade 1.5, lower protect (0.33 or less)
Force fp32 enabled
It could be that your mic isn't loud enough (raise input volume, add gain could fix it)
but I wouldn't rule out that the voice model isn't good enough.

#

Apparently to avoid such things, it needs a lot of sample data (more than 10 or 15 minutes of it at least).
The more samples it has the less these things happen

sullen heron
sullen heron
void flume
#

tg-dev's fork is bugged. I'm not too sure what is causing this, but- .. I mean we can try to optimize at least. I recommend inspecting the settings.json file. It might be that your sample rate isn't set to 48000 (when using server mode at least); you can tell from that.

sullen heron
void flume
#

the file is called "stored_settings"

sullen heron
#

Ah okay

sullen heron
void flume
#

Yeah

void flume
#

this guide helps with prepping the dataset for voice model training

#

as you can see there, for a quality model they diffidently go towards the 40~45 minutes or higher

sullen heron
#

Ah okay, although I'm quite content with my model currently as its only faced this problem on tg develop's fork

void flume
#

I recall reading somewhere that this is because the more data it has, the less it will make mistakes with stitching the chunks

void flume
#

When people speak (for the voice changer) too softly then these bugs can also happen; that is if the voice model isn't trained for whispers or soft sounds.

#

so adjusting the input volume or mic's hardware volume (or the one in windows) can help there

#

unless you're already doing that

sullen heron
#

Yes, I am

#

My voice model is also specifically trained with soft talking and whispering

#

And this problem happens whether I am whispering or not so

void flume
#

From what I understand is that it ... just happens on w-okada. It happens less on vonovox, which is why people tend to recommend that for the latest nvidia cards.

#

so there is diffidently a tool aspect in it there

sullen heron
void flume
#

The chunk / block size needs to be in the green; if you have all that your settings should be optimal

sullen heron
#

Like this?

void flume
#

That is way too much, but yeah basically

#

It's like ... 30 over what it actually uses when it starts to be green.

sullen heron
#

Oh

void flume
#

It's to keep in account the spikes that happen every now and then

sullen heron
#

Yeah there have been a lot of spikes

void flume
#

lower chunk size means that your GPU will be busier, but it should be fine

#

rmpve onnx seems to be made specifically for the dml version of the original rvc project's voice changer, which w-okada implemented. For two years, no one has updated it. I am not sure the regular one also has dml support, but it seems at least people are copy and pasting from eachother's projects rather than actually checking the code.

void flume
#

at least I can tell that from the discord chat messages.

#

I suppose it's an okay feature request lol

#

but yeah, idk if I can help with this further. ONNX models themselves are broken on the latest version of tg-develop's fork.

#

(related thought: maybe one version earlier has better performance for onnx specifically)

sullen heron
granite cedar
#

Hey, so can this Qwen3 TTS model be used for Speech2Speech as well?

sullen heron
#

It's not choppy or anything but the voice quality and whispering sounds pretty bad relative to tg-develop

sullen heron
hardy yew
#

Still there in 1.6.9

sullen heron
#

I was using beta when i was having that problem

sullen heron
hardy yew
analog obsidian
sullen heron
limpid jackal
#

hey guys is there a rvc other than applio?

sullen heron
limpid jackal
sullen heron
#

I don’t know how complicated applio is but i’d say yeah, its pretty complicated

#

the ui is in chinese (which you can translate to english) but i made my first model there

limpid jackal
#

i just want to make a cover 😭

swift peak
#

-rt

patent trellisBOT
# swift peak -rt
💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

sullen heron
hallow thistle
hallow thistle
sullen heron
#

And also do I really have to slice my data into 10 second audio clips? Cause many people say you have to and many people say you don't

analog obsidian
rugged wren
#

I don’t know how to explain it and I don’t know if this is how it’s meant to work but every single voice thing I have just changes my pitch and whatever, if I tune them they sound pretty much identical, why do they all sound the same?

#

When I’m home I’ll give examples but they all just sound like my voice just slightly different between them all

#

But I don’t know if that’s how it’s meant to work or it’s just my voice orrrr I’m doing something wrong

brittle wing
#

So I recently started using Applio and I've struggled removing backing vocals from my songs and vocals, UVR isn't working for me for some reason and at the moment I'm stuck using songs that don't have any backing vocals because that's all that works, I really want to make covers of all my favourite Japanese songs but I can't, can anyone help me out?

sullen heron
winged star
#

-rvc

patent trellisBOT
winged star
#

It's a 2h mp3

#

How much time is recommended to get good result ?

viral mason
#

It's better to use uvr locally if you have large data like that

#

Ask someone like a helper as idk anything about local uvr

#

Btw it's better to use wav over mp3

winged star
#

how much time you need of the voice ?

#

And what option I choose ?

image.png
#

I stay on music and vocals ?

winged star
polar lance
#

Hey guys. I have a question about training your own AI voice models (which sing).
I saw a guide that said you need to make a lot of individual vocal pieces that are up to 10 seconds long in FL Studio. Now, my question is
How many of these pieces do I need to make to get a decent voice?
Please help me if you know.

viral mason
viral mason
#

Just get around maybe 15-30 minutes of audio from anywhere as long as it's not too low quality, remove background music/noise and reverb/echo

#

Use a site like mvsep or minus x

#

Or uvr

polar lance
viral mason
#

Can't right now as I'm busy working with my grandpa

winged star
polar lance
dapper halo
#

RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Is this because the 5th gen gpus arent supported?

viral mason
viral mason
sullen heron
fast scarab
#

Hey, I’m trying to make a believable AI voice for music with RVC, I already have the vocals and the beat but I’m not really sure how to combine everything the right way, is there a good step-by-step tutorial that explains the full process?

viral mason
viral mason
#

Just keep it all one audio file

sullen heron
#

Well, at least now i know i dont need to, thanks a lot!

sullen heron
#

Does applio let me do that?

fast scarab
viral mason
#

-rvc

patent trellisBOT
viral mason
#

The docs for applio is the second link

viral mason
rugged wren
void flume
void flume
#

Red AI Hub docs for help with that (see bot message).

viral mason
#

u can find both vonovox and tg fork here

#

-rt

patent trellisBOT
# viral mason -rt
💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

subtle crater
#

The wokada application won't work. I can't get it to play game audio and playback the voice for me while using it. Yet on my laptop it works just fine.
I literally copied all my audio settings for input and output on my desktop so it matches my laptop

#

Could it be because I have a different virtual audio cable on my pc vs my laptop

void flume
# subtle crater The wokada application won't work. I can't get it to play game audio and playbac...
  • grab tg-develop's fork of w-okada instead of the official one. (see the above message for info)
  • use "virtual audio cable lite" (vaclite) instead of VB-Audio
  • Ensure your headset is the default audio output device, and your cable the default input device. (this will make sense later)
  • Set Input Device on the voice changer to your mic
  • Set Output device to the cable.
    ...
    Assuming other applications use the default device as mic, others should be hearing the voice changer now.
hallow thistle
hallow thistle
subtle crater
hallow thistle
fringe cliff
#

the weight site is shutdown how do I download new models

pale lotus
#

Hello

#

Mujhe 1 story do

pure field
#

hey does loading files on vonovox work for you guys? it says failed to load file for me

sullen heron
pure field
viral mason
pure field
#

i don't care about latency or lag because i only plan to run files through vonovox, with this in mind which settings should i max to optimise quality over speed

sullen heron
fringe cliff
viral mason
#

Those are specifically if they have weights links I guess

#

Just use huggingface

fringe cliff
#

okayy

rigid dove
#

"-rt"

#

Found the guide

viral mason
#

Like this lol

#

-rt

patent trellisBOT
# viral mason -rt
💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

viral mason
#

The " " was to show you what to type

cosmic epoch
#

How can i get a permission to post on the "voice models" section?

viral mason
cosmic epoch
rigid dove
#

Yeah it took me to a guide that I'm going through right now

viral mason
fringe cliff
#

Mods in ai help forum theres a guy proposing some job with a site without telling what it is

viral mason
thorn mantle
#

how do i make a voice model like can anyone send me a link?? the sapphire bot isnt helpăful cause i cant get to the google colabs or wherever i need to

viral mason
#

And also a site like mvsep or minus x, or UVR

#

-rvc

patent trellisBOT
viral mason
#

Here's the docs for applio

thorn mantle
# patent trellis

i went to thr uhh how to make voice models tab but i cant find any link im rlly slow

#

actually nvm i managed to fimd it

viral mason
#

👍

vagrant marsh
#

Does anyone have an applio or an rvc where it contains voice isolation (option where you separate the main voice from the background ones)?

hallow thistle
sullen heron
#

Hey guys, do any of you know if this is objectively the model overtraining?

If it is, which part of the graph do I test to find the best sounding epoch?

image.png
viral mason
vapid gust
#

why tf does my model sound like this

#

i used studio recording vocals???

alpine lotus
# vapid gust

either you trained the model without a pretrain or you used a wrong embedder while inferencing

abstract comet
hardy yew
hallow thistle
#

"The RVC/UVR5 combination fork" sounds possible but there are potential few limitations, likely the limited choice to select a UVR model within the same RVC/UVR5 environment, and how both softwares interact each other as automated as possible.

#

For better audio stem quality, UVR5 or "MSST" might be better choices.

dreamy willow
#

i have no idea if i am in the right place or nah BM_CatSit but when i run the http it just pops up, and shits itself BM_CatSit nothing happens (AMD 9060xt) (i'm sure i am not the first to ever ask something like this but BM_CatSit yeah i'd love some help)

dapper imp
#

Can anyone download a model from weights and upload it to hugging face or somewhere?

#

I want a model from there

hallow thistle
dreamy willow
hallow thistle
hallow thistle
dapper imp
#

If you can upload them, I can download it as fast as possible

dreamy willow
#

(?)

dapper imp
hallow thistle
dapper imp
#

But can you download that model and send it to me

hallow thistle
dreamy willow
hallow thistle
dreamy willow
#

I'Ll get to it and see what i can put under the covers WICKED

#

so...

#

nothing happened BM_CatSit

#

i have the main file location in one folder, i have the one linked in a COMPLETELY different folder aaand the issue is the same

hallow thistle
#

Try "Virtual Audio Cable lite" instead of the VB-Cable, see and hear if it fixes.

dreamy willow
#

well the main issue is that i can't even hear anything, realisticly speaking i don't even get an application to open up

hallow thistle
#

Don't try to run W-Okada fork with the "start_http.bat" batch file, it would mess things up.

dreamy willow
#

i didn't BM_CatSit i tried to run the main one from hujggingface

hallow thistle
#

This is the program file for W-Okada fork.

image.png
#

If you downloaded the "main" one from Hugging Face, it won't be that "W-Okada fork". Tg Develop hasn't hosted their repository on Hugging Face, only GitHub.

dreamy willow
#

so where the hell am i when i foujd like a huge lot of MMVCS version batch where i could download loads of version of it? 😭 on hugging face, i don't know if it's okada fork or whatsit'

#

like all of the "videos" i might find about it is listed that website

hallow thistle
dreamy willow
#

dude

#

yeah i saw it

#

so this is the single thing i might need riiight?

hallow thistle
dreamy willow
#

well thanks for the help I'll try to ask around some friendos too maybe they know something

hallow thistle
vagrant marsh
rigid dove
#

Is there a way that I would be able to hear myself while using the voice changer

viral mason
#

just set it as whatever headphones or headset u have

rigid dove
#

vonovox

#

Is the one I'm using

#

And I unfortunately don't see the toggle

viral mason
kindred kernel
#

So my Rvc Client isnt starting

viral mason
viral mason
#

Might have smth old

kindred kernel
#

MMVCServerSIO

viral mason
#

Uh

#

Could u send a screenshot maybe?

kindred kernel
#

of the link?

viral mason
#

The voice changer program you're using

#

Mmvcserversio is just the exe file that starts the program

kindred kernel
#

uhhh

#

thats the thing it wont start

#

let me try and find a sc

viral mason
#

Many have it so I wouldn't know which one you're talking about

kindred kernel
#

i have been using it for like 4 years now

#

now its fucking up

#

hold

viral mason
#

Oh

#

Yeah that's probably really old then

#

What gpu do u have

kindred kernel
#

Nvidia Geforce RTX 4080 Super

#

i dont have image perms

#

its just the default one

#

the one with white and blue

#

where it says realtime voice changer client

viral mason
#

Oh

kindred kernel
#

i dmed you

#

i got a new audio interface

#

after that it stopped working

#

if that narrows it down

viral mason
#

I personally use tg fork, it's pretty nice

#

-rt

patent trellisBOT
# viral mason -rt
💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Vonovox

A Realtime Voice Changer with similar performance to Wokada Tg-Develop Fork, with extra features, but it supports only Nvidia GPUs on Windows 10/11 unlike other options that have wider support. and without cloud options GUIDE

• Wokada Tg-Develop Fork

A personal fork (modified version) of Wokada Deiteris Fork, it just adds some Quality of Life improvements to it like supporting Spin Embedder and Audio Effects. Don't expect too much about it since the creator made it originally as a personal project. GUIDE

• Wokada Deiteris Fork

Deiteris' fork (modified version) of wokada that doesn't get updates anymore. GUIDE

⚔️ Wokada Tg-Develop Fork vs Vonovox

For Windows Nvidia, Both Wokada Tg-Develop fork and Vonovox have similar performance & quality. Users should read the pros and cons for both and choose based on their differences, such as UI and Vonovox's paid effects.
Read Wokada Tg-Develop Fork Pros&Cons & Vonovox Pros&Cons

⛔ Outdated/Discouraged

These options are not recommended for use.

• Original Wokada

Not suggested, older versions in youtube tuts are even way worse. GUIDE

• RVC GUI Mainline Realtime

The program is worse compared to the ones above, and much less updated. GUIDE

viral mason
#

It's the second guide

#

You'll need to download the 001 and 002 zip

kindred kernel
#

but that means i gotta put all the models back in one by one?

#

why did it stop working?

#

weird.

viral mason
viral mason
kindred kernel
#

i kept using that because it was the most robust version

viral mason
#

The screenshot u sent doesn't have any models u added tho, just the default ones that came with that app

kindred kernel
#

becaues i cannt START THE SOFTWARE

#

XD

viral mason
#

O lol

kindred kernel
#

to sc

#

it shows me the cmd

#

then closes

viral mason
#

Weird

kindred kernel
#

the console

#

i downloaded Wokada Deiteris Fork earlier and that one was such a mess

#

it didnt even keep the original pitch of the input?

#

lol

viral mason
kindred kernel
#

if the input sings a note the ai doesnt even sing the same note

#

🤦‍♂️

viral mason
#

like reverb and background vocals

#

btw since u wanna do that too u should get vonovox instead of tg fork

#

for whatever reason at least from what I know Wokada tg fork doesn't allow inputting audio files

kindred kernel
#

it used to work with the old one but these new forks keep changing the pitch?? im trying the one you told me to download rn

viral mason
kindred kernel
#

nah the second one

#

tg develop

viral mason
#

oh

#

that one in particular won't allow inputting song vocals

#

from what i know at least I cannot find it, that's why I said switch to vonovox instead

kindred kernel
#

nah it does when you send the music as your input

#

thats how i always did it

viral mason