✨│ai-help

[low shard]

the latest original has similar improvements (that got way later after they were firstly in deiteris fork), but we still suggest the deiteris fork for not having those useless beatrice v2 models, and having also force fp32 mode which helps with better quality and stable model

btw, it's up to date

[fervent trench]

I see

[low shard]

latest commit was last month

[fervent trench]

they seemed to have released a v2

[low shard]

nice btw

[fervent trench]

lmao why is it launching in the web

[low shard]

you mean the original wokada? yes I was talking about that in #✨│ai-help message , the latest wokada v2 beta has similar improvements after deiteris fork had them before, but we still suggest the deiteris fork for not having beatrice v2 and having force fp32 mode

[low shard]

please check https://rentry.co/forkvoicechangerguide#why-does-it-run-in-a-browser-and-not-its-own-window

[fervent trench]

fair

[low shard]

both original wokada and wokada deiteris fork use javascript & typescript

just that the original wokada made it's own "window browser" to open the local host url

[fervent trench]

the UI would regularly crash like a year ago

[low shard]

you can also check the github repo if u want https://github.com/deiteris/voice-changer, here u see that the 2 languages I mentioned are used much

[low shard]

yup, the wokada deiteris fork is easier to fix, it just has some issues with very specific browser like operagx

javascript moment

[fervent trench]

chrome is the only good browser, if you want everything to work

do you have some voice samples I can use to try the AI on

[low shard]

well yeah it's the most common browser so ofc sites make sure it works on that

[fervent trench]

some mp3

[low shard]

huh? wokada is meant for realtime btw

[fervent trench]

I know

but I'd have to talk

its midnight

and I have open back headphones, so it'd loop around

[low shard]

oh lol, i remember @crude flame sharing a speech.wav file before but forgot it

[crude flame]

whut

[low shard]

do u have any wav files u use for testing your models? I remember you sending like a f_speech.wav when I asked u that

[eternal aurora]

@low shard say i wanna make an ai voice the character is a robot how easy would it be

[crude flame]

oh yeah

that?

[low shard]

well ofcourse it's more complex, well since you got an igpu, you'd need to use cloud for that too

Train (make) RVC Models on cloud:

1. Prepare the Dataset
2. Setup RVC:
   Choose a cloud way to use RVC,

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Applio (ui)
  • Mainline (UI)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Mainline (UI)
  • Applio by Vidal (UI)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  • Applio (UI)
  • Mainline (UI, No guide as of right now)

3. Be sure to know about the tensorboard

Google Colab = Easier but risk of getting disconnected
Kaggle = Harder but way more gpu time
If you are looking for the easiest way and for free, is using https://weights.com/ which ofc uses RVC

RVC Inference (use models) on pre-recorded audio on Cloud

You can use either:

• Weights.com: Easiest Possible Ever Automatic
• Ilaria RVC Zero: Fastest free on cloud
• Applio UI Colab: RVC Fork with some extra features like TTS
• RVC AI Cover Maker UI: Automatically Separates the Vocals and Instrumentals, converts the voice and mixes them back

mainline is currently having issues though

[eternal aurora]

oh right thats true

[low shard]

yup thx

@fervent trench maybe try this

[eternal aurora]

im tryn replicate this

https://www.youtube.com/watch?v=0byUn5TQ2_Y

[fervent trench]

I will

[low shard]

well, it's better to have a larger datset

[eternal aurora]

ughh its a bit harder theres a 3 min video but theres another character speaking

[low shard]

welp

i mean u can try that short dataset, just don't expect too much

also remember u gotta clean it

i sent you all the links u need

[eternal aurora]

yeahh im checking it out rn

[eternal aurora]

yhh thats ture

tru

true

[brittle wing]

Intel(R) UHD Graphics

[low shard]

that's integrated graphics, pretty weak

do you have any other gpu?

check in task manager, performance tab

[brittle wing]

i have this one NVIDIA GeForce RTX 3050 6GB Laptop GPU

[low shard]

yeah that's better

I'm guessing realtime voice changer for calls?

[brittle wing]

yeah, just for fun

@low shard

[low shard]

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[grave fractal]

Hello, I'm new to the community. I want to ask if someone knows any AI that can change perspectives of an image, such as texts in it with their current fonts, background, etc...

[low shard]

1st link

wokada deiteris fork

[low shard]

To generate images for free, either:

• Use @brittle wing in https://discord.com/channels/1159260121998827560/1202754985255764060 (It's powered by DALLE3, from ChatGPT+), pretty easy
• Another easy and good ways with weighs.gg are:
  • Use /image with @earnest musk in https://discord.com/channels/1159260121998827560/1202754985255764060
  • Create an image on their site https://www.weights.gg/ (which you can also use LoRAs, Low-Rank Adaptations, basically a small trained additional model to adjust your generation)
• Use Open Source Models like stable diffusion & flux that could be a bit **harder **but good, what's ur pc gpu? As you could run them locally (on ur pc) or on cloud (remote good pc)

img2img seems like what you're looking for, maybe try one of the things I said above

[earnest musk]

:wave: @low shard, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[grave fractal]

Farewell, I didn't ask if someone knows an AI that can generate images... I asked if an AI can change perspectives of an images, such as texts with the same fonts in it, as well with their background, etc...

[fervent trench]

which one do I use again for gtx1660 ti

[odd shale]

Use rmvpe with 256 chunk and 2.7 on extra

[fervent trench]

it sounds bad

lol

I need to be really close to my mic

[low shard]

I mean, img2img is a technique where you can input an image and then also change it based on your prompt and the parts you select

for example, the weights.com i said above could be one way to solve this with: https://www.weights.com/create/canvas

[low shard]

could you also share a screenshot of ur whole wokada?

[eternal aurora]

ngl im doing something which is tedious asf

with that 3 min clip with the charcters speaking im just extracting audio when the robot speaks

[odd shale]

If it sounds bad, it can be also a problem on the model itself or your own mic

[fervent trench]

I need to talk rather loud for it to pick it up properly

oh it probably was the input sensitivity

[low shard]

set the in to a bit higher value

[eternal aurora]

how much seconds of audio do i need

[low shard]

I mean there isn't a rough limit built in rvc, but 5 mins would be usually the minimum for smt decent

[swift thunder]

How can I download the pre-zoned pretrainings in Applio Colab? The cell that was there before doesn't exist.

[eternal aurora]

ah okay

[upbeat mountain]

i need help wwith a good girl voice changer too troll on omentv i got gtx 4060

[crystal gull]

Thanks, I've switched over to LM Studio and this is quite swell. GPU works perfectly, and it's got some nice features and configuration. Simple to edit. Haven't touched the api yet but I'm sure it'll work swell based on the experience so far

[low shard]

i think you meant rtx 4060

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

1st link

wokada deiteris fork

[upbeat mountain]

oh yeah

do i needd to install them things?

[low shard]

yep, read up the guide

[eternal aurora]

@low shard imagine I found a 12 min vid of him speaking

But there’s music in the bg 💔💔💔

[odd shale]

You can easily isolate the speech.

[eternal aurora]

Can I?

How do I do that?

[odd shale]

Yup.

[odd shale]

https://huggingface.co/spaces/TheStinger/UVR5_UI

Use this Huggingface space.

[eternal aurora]

Yep

[odd shale]

I personally suggest you to use the Gabox FV4 model.

[eternal aurora]

okay willdo

[upbeat mountain]

i need help with the voice chagner like i wann sound like a girl but everyone i try its shit

[odd shale]

Also drag the audio to where it asks you to, wait until it loads completely, select an output format and click on "Separate"

[eternal aurora]

okayy ill lyk how it goes

[odd shale]

Maybe these docs will help you.

https://rentry.co/ForkVoiceChangerGuide

[simple ore]

cool

[upbeat mountain]

i was sent that but i dont understnad what im meant to do

[odd shale]

Just read the guide and don't skip anything.

Also make sure to download and use deiteris fork instead of OG W-Okada, and lastly, it's matter of testing which model from #1175430844685484042 channel works better with your voice and settings

[eternal aurora]

im getting this error

[odd shale]

That error pops out when trying to access the UVR HF space?

[eternal aurora]

yeahh the wokada thing

[odd shale]

Or are you trying to do something else?

[eternal aurora]

i press the link that it gives me and it brings me to this

[odd shale]

Yeah no, it seems you're trying to do something else.

That HF space isn't for realtime.

It's for stem isolation

[eternal aurora]

how do i fix that tho?

[odd shale]

I don't know, but if you're trying to use realtime, what's your GPU?

[eternal aurora]

im using it on cloud

[odd shale]

But what's your GPU

[eternal aurora]

so its the clouds gpu

[eternal aurora]

havent got one

[odd shale]

I see.

[odd shale]

I'm not sure but, did you put your Ngrok token correctly on the notebook?

Or did you ran out of time usage?

[eternal aurora]

yhyh i did it was working fine before

[upbeat mountain]

anyone got good female pitches

[odd shale]

Then i got no idea sadly.

[eternal aurora]

time usage is 30hrs and i started like 2 hrs ago so no way

[eternal aurora]

okay thats np

[odd shale]

Nope, it's matter of testing which pitch fits the best with your voice.

Altho you can test with 13-15 or 17 on pitch

But will depend on the model and your own voice though

[eternal aurora]

i fixed it

[royal juniper]

hey guys is there a free voice changer that I can use without training any voices? like something I can download and use and preferably without using someone else's voice?

rtx 4060 ti 16gb

i310100f

16gb ram

I want a voice changer for live streaming

ai voice changer while I'm live.

[low shard]

i mean, you could use wokada deiteris fork and use premade RVC models made by other users, or are you looking for not using any models at all?

[royal juniper]

I wanna use models. I just don't wanna train them myself

[low shard]

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

1st link, read up the wokada deiteris fork guide

[royal juniper]

what is that?

[low shard]

@royal juniper though, the models trained by other users, will be trained by some voices somewhere ofcourse, like homer or peter griffin's voice

AI has to be trained on something

[royal juniper]

I can mix voices trained by others right?

to create an original voice? will it have any quality problems?

[low shard]

wokada is a program designed to run RVC (speech to speech) models in realtime

there's 2 versions, the original and the deiteris fork

currently the deiteris fork is the most suggested one

[low shard]

iirc yeah, the merge lab should work all fine

else you could try something like voice mod, which isn't really AI but just voice effects, though that has a paid tier

[royal juniper]

I know w-okada. but what's the difference with deiteris fork?

I honestly want to avoid training voices myself because it's too much effort. ai hub is safe for downloading voices right?

like if I mix 2 or 3 voices it'll be completely new voice Right?

[low shard]

the deiteris fork is a modified version by deiteris,

it was always used because it has improved quality and performance, though the latest original wokada performance comes close to it, the deiteris fork is still suggested for not having beatrice v2 models, and also for having a force fp32 mode, which increases model quality and stability, along with having a more stable ui

[low shard]

yeah, wokada deiteris fork also makes sure it only loades the model and nothing more

for more technical info there's: https://rentry.co/forkvoicechangerguide#is-wokada-deiteris-fork-safe-are-the-rvc-models-safe

[royal juniper]

ok thank you for today.

I'll try this tomorrow with my pc

[low shard]

alright, yw and lmk

[eternal aurora]

Ngl my shit is lagging

And my perf is red tf

[low shard]

send a screenshot of ur wokada

[low shard]

because it's running on cpu

set the gpu to the T4

[eternal aurora]

oh shu

my thing was bugging out before so i had to reset it

[low shard]

ohh, changing that should fix it

[eternal aurora]

yeah it does tysm

i love you

[eternal aurora]

is this fine?

@odd shale im unsure with what to do with the audio separation link

i changed the model

and uploaded the link

nvm i sorted it

no way its separated

[brittle wing]

-colab

[patent trellis]

📒 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

[brittle wing]

how do I decrease the delay?

for like voices

[eternal aurora]

reduce the chunk size i think

[brittle wing]

its not letting me, like i click it and nothing happens

[eternal aurora]

are you still playing the voice>?

make sure u press stop before you chnage it

[brittle wing]

thank you!

[eternal aurora]

np

[amber merlin]

Hey! im still having issues getting the software to recognise my GPU

[brittle wing]

Hi, can I finally get an answer, please?
I'm planning on training 2 models, how do I resume in Applio No UI colab when I switch accounts cause my dataset is big & one account isn't enough

Can finally someone tell me what cells do I run, the sequence and stuff

[hallow thistle]

W-Okada, RVC or Stable Diffusion?

[crystal gull]

You have no previous messages here, are you sure you are in the right server?

[drifting night]

torch\nn\utils\weight_norm.py:28: UserWarning: torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.
warnings.warn("torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.")

[knotty moth]

it's not an error message and can be ignored if it works as is

[hallow thistle]

The "UserWarning" is a warning about certain code lines being deprecated and can be changed as soon in newer PyTorch versions. The program can continue to work anyway.

The actual error starts with something like "Traceback (most recent call last):", then follows with paths to related files, and then the error message.

[long oyster]

i get some trouble with vac lite

i can't hear my voice at mic test in discord

[hallow thistle]

You can set "monitor" output to your speaker on W-Okada to hear the audio. Other than this, you can restart the whole W-Okada or try another browser.

Also, Sup1 is a deprecated noise gate algorithm, try turn Sup2 on and increase In. Sens. number until there's no noise in the background.

[long oyster]

hmm okay

[long oyster]

my setting is good right? or should i change anything? i don't know about KLM, epochs and steps like i can't find it

line 1 can't work in discord and valorant

yesterday i download vac lite and it's use normal

but today i can't chat

and don't see line 1 in valorant

can somebody help me?

[knotty moth]

double check settings like this, also try another model, perhaps including older ones

[thorny cargo]

Hey y'all, around late 2023 or early 2024 I remember installing a local gui that could allow me to train voice models being the pth and index however since then I am on a different device and I don't remember where to install it. Under the Guides tab which one contains the local model trainers?

[knotty moth]

training locally on devices other than pc? no way!

[thorny cargo]

laptop vs desktop i guess

but yeah

Anyway yeah where is it, I remember there being a place to downlaod it

[knotty moth]

RTX 4060 laptop or any 6+ GB vram is minimum recommended to do so

[thorny cargo]

all good

damn it

[knotty moth]

that's more than enough but you'd worry on laptop heating all the time

[thorny cargo]

I also have that covered you wouldn't believe me

the most redneck ass cooling setup for this poor thing

but yeah

I just need to find the trainer

[knotty moth]

also with budget for laptops with that similar spec you could afford a decent 4080/super desktop build

[thorny cargo]

You see I do have a desktop with a 4090, but I travel and can't take it with me everywhere because that would be an honest to god death wish

It's already a ticking time bomb with the connector shenanigains imagine travelling with it

[knotty moth]

remote desktop is the viable solution unless for gaming

[thorny cargo]

I am away from it for 5 months

Not gonna risk keeping it plugged in

but I do get the sentiment

[knotty moth]

to save the electricity bill huh

[thorny cargo]

nah I can't check the cable or clean out the fans, can't maintenance incase anything happens

its all around a risk

I rather unplug everything and stow it away

[thorny cargo]

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

I assume this may be it

wait

no

its not from scratch

agh, damn

[knotty moth]

grab that vacant domain for your website, or check out the docs here https://docs.aihub.gg/rvc/local/applio/

[thorny cargo]

this isn't the exact one but it most certainly checks off all the boxes of what I needed

much appreciated

this is what i get for not taking my boot drive before travelling

have to scavenge for everything I didn't have

[knotty moth]

if I were doing travel all the time, I'd rather give up on doing workloads using local desktop/workstation

[thorny cargo]

If it was 1 month i'd be like

sure

I can take a break

but for 5 months I'd probably want to mess around with the tech a bit here and there

I got this laptop off FB for a great bargain so I didn't initially plan to do this

[low shard]

Elaborate more

[brittle wing]

im using a 4060, the problem is my normal mic without the voice changer is normal like it doesnt capture anyones voice, but when i use the voice changer it also changes whatever noise is coming on the pc up

[low shard]

elaborate also the rest, send a screenshot of ur wokada

!give-media-perms 1h @brittle wing

[eternal aurora]

@low shard

My ai voice has an accent but I can’t seem to hear it even when the index is set high

[low shard]

did u use the tensorboard

[eternal aurora]

No, what’s that?

[lofty lichen]

what did u use to remove

plugins only thing that works for reverb/echo

[low shard]

https://docs.aihub.gg/rvc/resources/training/#tensorboard

Did you just train it to 500 epochs

Bc there isn't a right amount of epochs

[latent yacht]

do you guys know any video Upscaler that I can use on my PC locally? ( I have an Nvidia RTX 3060Ti ) ( also ping me if you do respond to this )

yeah

[eternal aurora]

Wait 1 sec, is this for a pre-made model or the one I’m going to make?

[lofty lichen]

i use acon deverb 3 + dxrevivepro for talking echo

most of the time

[low shard]

Did you train a model or just used one made by someone's else?

[eternal aurora]

i used one that someone else made

[low shard]

ohh nvm then, I thought you were making one

well it's better you just try another model

[eternal aurora]

oh okay

also how long will it take to make it?

i have done the first step

https://docs.aihub.gg/rvc/resources/dataset-isolation/

im on the traning part

[vestal helm]

https://tenor.com/view/black-guy-what-gif-19134356

for

models??

[olive plover]

Hello support

[brittle wing]

Ok how do i fix this error in the run-applio.bat file after i changed the theme oof the app

[olive plover]

Is there a new update for fork??

Or when will an update be made to it because the last update was in 2024

FORK

[low shard]

could you elaborate:

• your pc gpu
• what did you do exactly
• what's the error
• a screenshot of the issue

!give-media-perms 1h @brittle wing

[brittle wing]

No ican't send i'm not in verified

[low shard]

it could take some hours for training a model

[low shard]

I gave you image perms, you should be able now

maybe retry

[brittle wing]

wait

this one

[low shard]

the last update for wokada deiteris fork was the b2332 one from december 2024, there isnt any newer stable releases yet, also it would be better next time you specify the program since fork just means modified version in IT terms

[low shard]

what team did you choose? what applio version are u using? and what's your pc gpu?

[olive plover]

Do you mean that a newer version of b2332 will not be released?

[brittle wing]

the bugfixed one

My pc GPU is Nvidia Geforce RTX 3050

[low shard]

I didn't say that the program won't be updated anymore, I just said that the latest release of right now is b2332, we don't know when a newer stable release will be out, it's better to ask deiteris ( @wispy lodge )

[low shard]

3.2.8-bugfix ?

and what theme did you change it to?

[brittle wing]

ye

the applio one

[olive plover]

Hello @wispy lodge

Is there a update copy of b2332 for FORK, and if yes, when will it be?

[brittle wing]

no message yet

[low shard]

could you please try to reinstall the program?

[brittle wing]

oh yes that's it

[eternal aurora]

Ohh okay

[low shard]

tell me if it's fixed after doing that

[brittle wing]

My file is still downloading

i have the precompiled zip file

It still doesn't work

nvm it works now

[vestal helm]

remastering

like clean up?

what’s the process? @modern marlin

ahh

Okay that’s cool

[low shard]

great, let me know for further issues

[brittle wing]

No issues found

[low shard]

great

[eternal aurora]

@low shard which one should i use?

my character is a robot

[low shard]

it's better you just use the original pretrain in this case

there's also https://discord.com/channels/1159260121998827560/1242833862640402464 but it's trained on only 2 hours

[eternal aurora]

ohh okay

[eternal aurora]

so the v1?

[low shard]

i'm not talking about the train architecture, i'm talking about the original default pretrain you already have without needing to use ones made by the community

[eternal aurora]

ohh okay where do i go to get that?

[low shard]

it's already inside of mainline rvc or applio

just by training normally

[eternal aurora]

okay ill see

sorry where is the pretrain tab?

[low shard]

you don't need to download any pretrain

the default one is already set

[eternal aurora]

okay, but wheres that?

[low shard]

what rvc program are you using? applio ui colab?

[eternal aurora]

yep

[low shard]

huh what are you installing locally?

you said you have an igpu, meaning too weak for RVC

that isn't applio ui colab

[eternal aurora]

tensorboard

[eternal aurora]

yep

[eternal aurora]

oh is it not?

[low shard]

on applio it's already installed

and you are on cloud

meaning you don't need to install anything

the tensorboard will be opened directly when you start the run ui cell in the applio colab

right below the public ui

[low shard]

@crude flame maybe this should be should be said in a more noticeable manner in the docs

[eternal aurora]

ohh okay

[low shard]

you can just delete what you just downloaded

[eternal aurora]

okay will do

this is gonna sound stupid but where do i go to access appolio

[low shard]

it's called Applio, it's an RVC fork

you got some cloud options:

• Kaggle: 30 hours weekly but needs a phone number and is a bit harder
• Colab: max 4 hours of daily gpu but could be even less

[low shard]

don't this is for locally, your pc is too weak

do u want kaggle or colab?

both are listed in https://docs.aihub.gg

https://docs.aihub.gg/rvc/cloud/applio-colab/ and https://docs.aihub.gg/rvc/cloud/applio-kaggle/

you can choose either

kaggle is more suggested for training

[spice nacelle]

I need help from an expert regarding ai image generation if possible

[low shard]

what do you need exactly?

[spice nacelle]

can i share images here it would be easy to explain

[eternal aurora]

ohh okay##

[eternal aurora]

ill take kabble as i got it all swt up

[low shard]

!give-media-perms 40m @spice nacelle

[low shard]

alr, i sent the guide link

[spice nacelle]

im looking for a way to do this if anyone can help like how this person turned this image into a stunning thumbnail

[eternal aurora]

yeah thanks ill lyk later how it goes

[spice nacelle]

i know he used image to image but i have never done this before so i was wondering if anyone knows how and where i can do this and using what loras or models

[low shard]

• easy and good ways with weighs.com are:
  • Use /image with @earnest musk in https://discord.com/channels/1159260121998827560/1202754985255764060
  • Create an image on their site https://www.weightscom/ (which you can also use LoRAs, Low-Rank Adaptations, basically a small trained additional model to adjust your generation)
• Use Open Source Models like stable diffusion & flux that could be a bit **harder **but good, what's ur pc gpu? As you could run them locally (on ur pc) or on cloud (remote good pc)

[earnest musk]

:wave: @low shard, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[low shard]

it's for sure img2img, I gave you some platform options and also asked your pc gpu to check if you could do it locally with programs like ComfyUI

[simple ore]

something wrong with translations.. corrupted or incompletely unzipped

[brittle wing]

What I'm already using

[simple ore]

dont use python 3.13

90% of stuff does not work with it

[brittle wing]

What's the error in this audio

[silent stratus]

some models do that, try a longer input and then just cut it to the part u need

[brittle wing]

how do i get a better voice model so my model won't be rejected

[eternal aurora]

Yeah dw I turned it off

[silent stratus]

ohh I thought the model was quiet and u didn't like it

[brittle wing]

is the model good

[silent stratus]

ur data seems overcleaned and quiet

[brittle wing]

okay

[fast jackal]

how can i search ai voice models for mmvcsserver&

[cinder sable]

hi i have a problem when audio is not showing up on voice changer basically i cant set my headphones to it

[odd shale]

Do you mean w-okada?

You can check the #1175430844685484042 channels, there are plenty of models on that channel

[spice nacelle]

I am so sorry I didnt respond to your message sooner some came up so i had to leave but thank you so much for the info ill try it and see how the results are

[low shard]

elaborate:

• what's your pc gpu
• what tutorial link did you use
• a screenshot of the program
• whats the issue

[low shard]

elaborate:

• what's your pc gpu
• what tutorial link did you use

be sure u didn't use video tutorials, they are old

[low shard]

yw

[hallow thistle]

MMVCServerSIO is the folder/program name for W-Okada. For voice models, look in #1175430844685484042 or find one on Weights.com.

[short marsh]

when i talk into w okada the app just closes how do i fix this?

[low shard]

elaborate:

• your pc gpu
• what tutorial link did u use
• what u want to do

be sure to never use video tutorials, they are outdated

[short marsh]

my gpu is nvdia gtx 1050ti i used Modified W-Okada
W-Okada Fork Guide
and i also cannot see the box for my pref value

[low shard]

!give-media-perms

!give-media-perms 1h @short marsh

send a screenshot of ur wokada

[hallow thistle]

"Fork" W-Okada and "original" W-Okada are two different versions of W-Okada made by different authors.

The version in this screenshot is the original one.

[short marsh]

oh so how am i able to download the fork one and which one is better for my gpu?

[low shard]

you're using original wokada

uninstall the one you got along with vb audio cable

[low shard]

wokada deiteris fork

-realtime

[patent trellis]

💻 Local Realtime RVC

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

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[low shard]

read the 1st link

be sure to not use video tutorials

[short marsh]

how come i need to uninstall this?

[low shard]

you're using an old version of original wokada

[short marsh]

okay thanks alot

[low shard]

you deffo used a youtube tutorial, those are old for wokada and rvc

[low shard]

yw and lmk

[oblique glacier]

pls hlep

voice laggy

[crude flame]

make it more clear that you dont need to install the tensorboard if you use applio?

[low shard]

other than that, that installing tensorboard is only for local mainline users

[low shard]

set input to microphone, the voice doesn't seem laggy

is it laggy only during a game? if so, open that game, set the graphics to the minimum, close useless background programs, then show a screenshot of ur wokada

[oblique glacier]

no i have no game just google chrome

[low shard]

well the program doesn't seem lagging, your perf value is green and just 36ms, maybe you meant robotic?

[oblique glacier]

input is already at rvmpe

[oblique glacier]

not robotic more like

[low shard]

input should be your microphone, sorry i did a type lol

[oblique glacier]

old woman with broken voice

[analog obsidian]

for both mainline and applio the best way to choose epochs is to listen to them anyway, the graphs are there just to show you if there are any irregularities in the training

[oblique glacier]

alright

my graph is not normal

[low shard]

he's talking about something else

[oblique glacier]

ah ok

[low shard]

crackling?

[oblique glacier]

but still that indicates something

[oblique glacier]

i mean we can test it together if u can join

a vc

cuz idk how to

describe

[low shard]

no, this only indicares that you're having a 37ms delay, and it's good,
when the perf value it's good it means it's not lagging

[low shard]

okay wtf lmao

@oblique glacier have you tried playing with the pitch?

[oblique glacier]

i see

[low shard]

also, set the extra to 2.7

[oblique glacier]

how much do u wxant me to put it?

i put at 8 cuz girl voice

[oblique glacier]

i had it originally but i think somebody told me to put it at 2

maybe it was u i dont remember

or the whale guy

[low shard]

well there isn't a right amount

you gotta play with it till it sounds right because it varies depending on your own voice and also voice model

well still sounds a bit like whispering-cracking

[oblique glacier]

but it shouldnt make much of a difference no? if i change the pitch too much the voice model won't work like i want

that girly voice

[low shard]

@oblique glacier have you tried https://rentry.co/forkvoicechangerguide#discord-crackle-fix ? or also trying another voice model

[crude flame]

it also helps to speak like the voice

[oblique glacier]

i didn't try it yet

do u have a good voice model?

[low shard]

yeah, don't put it either too low nor too high

[oblique glacier]

lel i already have a high pitched voice as a man, embarassing

[oblique glacier]

is 5 good?

[low shard]

but for some models might be 10, some others might be 8, yk

[crude flame]

alright, time for bigger text

[low shard]

well, I can't tell you the perf value for your voice with every model

nvm it's just deeper now

[low shard]

maybe try that

[low shard]

https://rentry.co/forkvoicechangerguide#voice-models-to-try-out

[spice nacelle]

@low shard you my frined a a life saver

I just created this thanks to you

[oblique glacier]

nice image

LOL

[spice nacelle]

but for some reason i cant add loras in ther

[crude flame]

the heck LOL

[low shard]

yeah bc i just saw some applio colab users thinking they had to install tensorboard locally lol

[low shard]

youtube, thumbnails 😭

[spice nacelle]

and is there a limit to how many i can create

[low shard]

@oblique glacier have you tried one of the models i said or the discord crackle fix ?

bc i still hear you with a broken voice

[low shard]

huh you can, u just used a lora

those are loras

they are just commonly called image models in the site

[spice nacelle]

sorry i think its models i wan to say

[crude flame]

other than this and inst model recommendations what was the other change you wanted me to make? I forgot it and i cant seem to find it, if it even is a thing

[oblique glacier]

nope

discord crackle fix? havent heard of that

[spice nacelle]

can i add different models in there for a different style

[low shard]

please try the 2 fixes I told you

[low shard]

it's on the extra part of the guide

[spice nacelle]

is there a limit or anything like that for image generation on that site

[low shard]

you can use up to 3 loras at the same time iirc

[low shard]

iirc like 100 images a day in free tier

@oblique glacier A VAC (Virtual Audio Cable) makes a fake audio device, used to re-route the audio of different programs

In Wokada context, it's used to get the output of wokada as the input in other programs

[spice nacelle]

wait for real 100 images per day??

[low shard]

you installed vac lite right ?

[low shard]

https://www.weights.com/premium

@oblique glacier wait no

I asked you to do https://rentry.co/forkvoicechangerguide#discord-crackle-fix

@oblique glacier i think you confused with https://rentry.co/forkvoicechangerguide#press-enter-to-continue-failed-to-download-or-verify-files , you don't need to do the other extra steps, only the exact link I sent you

[low shard]

which is this^

@oblique glacier no nope that, don't follow the other help steps, they aren't unrelated, that help part is only for when your program doesn't start, but your program starts lol

yea 😭

u can

u can just ignore that drive thing u downloaded

@oblique glacier yeah still not good, this might be a model issue

it's better you just use another one

there isn't any cut

but it sounds weird

you could try some of the models mentioned in: https://rentry.co/forkvoicechangerguide#voice-models-to-try-out

[low shard]

@oblique glacier ^

@oblique glacier it sounded bad because you used the male sample on a girl model lol

not sure if they have a female one

@oblique glacier btw rename the files before uploading

bc all models on weights are named as 'model'

@oblique glacier yes, model is the pth and index is .index

[spice nacelle]

@low shard I'm so sorry to bother you again but can you cehck this image to see what kind of generator or lora is this person using to create such nice looking images IK he created these characters individually and then used photo shop to recreate these thumbnails but the characters are so nicely made

[low shard]

@oblique glacier put your input to microphone, then go on discord and be sure the input is line 1 here instead

did you close it?

open it, select the model, then click start

[low shard]

well I can't know what image gen model or lora they are exactly using, but might be flux (the model weights.com uses by default) with some anime lora

@oblique glacier yeah it's installing internal files bc u deleted them before accidentally mixing up the guide 😭

[oblique glacier]

😭

[low shard]

@oblique glacier could u type what u just said

[oblique glacier]

do i need to use vertical synchronization when i'm playing?

vsync

[low shard]

oh the vsync, well that helps with not having screen tearing in games

[oblique glacier]

yes but do i need to enable it for a smooth voice?

[low shard]

no that doesn't make the voice quality better, it's just related to the games only

@oblique glacier u sure u selected the right model? it sounds like a man

[oblique glacier]

😭😭😭😭😭😭😭😭😭😭😭

[low shard]

u have like 10 slots named "model" so not sure if u just selected the wrong one

yeah u selected the wrong one lol

[spice nacelle]

i also have some questions regarding this topic recently i tried cloning a voice using a 30 min clip inside all talk tts the voice can out fine but the way the person speaks isnt exactly what i was hoping for

[low shard]

@oblique glacier don't rename on the site, you have to rename the voice model files

the renaming models on the webui is bugged

[spice nacelle]

did i do something wrong or used the wrong software to clone the voice or should i use longer clips

[low shard]

it's better you rename the model files, then re-upload them

[low shard]

I never really used alltalk tts, but maybe u could check https://docs.aihub.gg/tts/tts-tools/

@oblique glacier i hear ur normal voice rn

@

it's better now, if you want to hear yourself u can also set monitor to headphones btw

u might be a bit lagging

yup

[spice nacelle]

which one do you recommend that is able to create long paragraphs of text into speech with minimum time

[low shard]

well, any other issues?

yeah i hear u good

[low shard]

11labs might be what you need, since it's on cloud and can be pretty fast, and also has a good quality

@oblique glacier yeah but not when u eat the microphone 😭

RVC has a limitation to laughing

what?

[oblique glacier]

nonchalant mode

xD

[low shard]

lol

also u might need a bit of voice acting

to make it sound realistic

[oblique glacier]

yeah i need to distort my voice a bit

[low shard]

like the model doesn't give emotion alone

[spice nacelle]

but its not free i have to create audio over 1 hr everyday and itll cost too much to generate even one audio that is over 1 hr

[low shard]

@oblique glacier well, seems like it's all fine now? i can get off the vc?

alright, yw @oblique glacier

[low shard]

ehh, you might want to try edge tts but it isn't emotional

or something like f5 tts

but unfortunately i dunno about alltalk

[spice nacelle]

alltalk was the fastest one that i found so far ive been searcing for a while now but ill give them a try and see if the results are any better

[low shard]

i mean if u only care about speed, pipertts is an option, but it isn't good quality

[spice nacelle]

can i send a video clip here to show what kind of result im looking for?

@low shard when you get the time please take a look and if possible I'm also looking for a pre trained model for this voice but I'm not sure what it's name is so if anyone knows please let me know I would appreciate it very much

[jovial pollen]

genuine question : on deiteris repo there is 001 and 002 to download, how can i make sure both are extracted together ?

[woven sun]

Do you know of any guide to download an AI to make images or videos ? I have an 8GB RTX 4060 Ti and I wanted to use it, I'm pretty new to this.

I tried to download SD but there were some problems with the PY files and I couldn't use it

[pastel oak]

Put both files into a folder (or keep it in downloads) -> right click 001 -> 7zip / winrar -> extract to ....
it will automatically detect the 002 and extract it

Or you can download the precompile i made, its on the guide

[jovial pollen]

ooooo

can u share the guid eagain pls

[pastel oak]

https://rentry.co/forkvoicechangerguide#download-nvidia-on-windows

[jovial pollen]

many thanks

[placid holly]

i checking the AMD version it directing to DML one, but when i checking the release in Github there is Amd64-cuda, what the different?

while i know for Nvidia is Nvidia-b2332

[analog obsidian]

amd64 is the cpu bit mode, not related to amd (the gpu)

its also often called x64

[low shard]

-# 32 bit programs are also known as x86 btw

[analog obsidian]

yuh

[low shard]

so you're trying to find a voice you don't know the name off? well, simple you cant

[low shard]

maybe comfyui or pinokio

[analog obsidian]

its called amd64 because amd made the first 64bit cpu

[low shard]

while intel made the first 32 bit cpu

[analog obsidian]

so the package in simple words says: voice changer, 64 bit system, dml

cuda for nvidia
dml for amd

[placid holly]

ah i apologize for my confusing question, now i understand

in Hugging face it named differently

i though that one is another version for AMD

Thank you for the Answer 👍

[simple ore]

technically the voice changer can be made to run on AMD with zluda instead of directml

[low shard]

would that have any benefits tho?

[analog obsidian]

less cpu usage

[simple ore]

directml speed is ass

[oblique glacier]

@low shard

can i stream with the voice changer?

i won't have any issue with it?

[placid holly]

yes, it just make the line 1 is one of your stream software audio output

just like in game

[oblique glacier]

yes ofc

ty

[placid holly]

but there is delay

[oblique glacier]

no problem, as long as my voice is neat then it's good for me

[drifting night]

Im having an issue
everything works including the default voices it comes with
but none of the custom voices i added work for some reason i dont know why i did everything right

[paper bloom]

hey

hey anyone knows a good voice changer currently im using wokada i but like in games and stuff its not working very good

i would like to share a screenshot i just made but idk how

[placid holly]

please tell me what the spec and setting you using

is it the converted voice or just the game when the problem exist?

[paper bloom]

idk like when i run a game it starts buggin

especially when i choose server instead of client

my gpu is nvidia geforce rtx 4060 (8GB)

is my gpu to bad maybe?

and its not the voices i think maybe my graphic card is to bad idk

[low shard]

what game? and put game graphics to the minimum possible

also, share a screenshot of ur wokada

!give-meida-perms 1h @paper bloom

!give-media-perms 1h @paper bloom

[paper bloom]

like marvel rivals

but also sometime in vr chat or phasmaphobia

[low shard]

lemme guess, you used a youtube tutorial?

[paper bloom]

yes^^

idk how that works im a noob

[low shard]

that answers your own question

all video tutorials are outdated lol

vb audio cable gives random issues on windows as reported by users

and you're running on an old version of original wokada, more than a year old

also the settings are awful

@paper bloom uninstall everything you got off the youtube tut (wokada and vb audio cable), completely forget about them

read the wokada deiteris fork https://rentry.co/forkvoicechangerguide

[analog obsidian]

bilibili w-okada tutorials are way better than youtube tutorials
but for some reason there they prefer fcpe over rmvpe

i mean i guess they like that its faster

but at least they're aware crepe sucks for realtime

[low shard]

bilibili has wokada tuts?

[analog obsidian]

yeah

tho they prefer to use mainline's realtime

[paper bloom]

is there a good tutorial i can use?

[low shard]

[low shard]

yes, I already gave you it: #✨│ai-help message

[analog obsidian]

i mean mainline realtime is better than original w-okada

[low shard]

wokada deiteris fork though

maybe u should also compare latest original wokada with mainline realtime lol

[analog obsidian]

good idea

[paper bloom]

is there no video turorial im rly a noob ^^

[low shard]

@paper bloom uninstall the ones you got, read the guide I sent, then remember to set marvel rivals graphics to the lowest possible

[analog obsidian]

no

sit and read

[low shard]

nope, video tutorials are more outdated, since they can't be easily edited

please, read it out and let us know for issues

[paper bloom]

so the problem might be the old version i have?

[low shard]

yes, the problem is you're using an ancient version

forget everything you got off that youtube tut

[analog obsidian]

technically the real problem is crepe being horrible for realtime inference
but besides that, the original w-okada is a buggy mess

this server doesn't support that anymore

so every helper/mod will tell you to download the fork

[low shard]

plus the ones used by yt tuts are over a year old, performance would be shit especially for games compared to the deiteris fork

[paper bloom]

can i still use the voice i used before for the new one atleast?

[analog obsidian]

change f0 to rmvpe

should improve perfomance

in the game be sure everything is at the lowest possible, res 1080p, 60fps cap

extra is cpu accelerated in the original version, that might cause a cpu bottleneck if your cpu is weaker than the i7-13th

[paper bloom]

my fps goes up to 200 in the game

[low shard]

yes, they both use RVC models

[analog obsidian]

yes but the nvidia driver chooses to give more priority to your game

so it focuses more on rendering those fps

rather than giving those resources to your ai gpu accelerated voice changer

[paper bloom]

oh okay can i cap it?

[analog obsidian]

60fps cap is good

but if you rlly hate 60fps, try 120fps

[paper bloom]

but how do i do it in the game settings?

[analog obsidian]

marvel rivals?

[paper bloom]

or in genreal desktop settings

in general how to cap fps for games

[low shard]

https://www.youtube.com/watch?v=V_OejS9F4PI

generally: just look at the game settings

[analog obsidian]

you can use the in game fps limiter or 3rd party limiters like riivatuner

[low shard]

most games already have built in settings to limit fps tho

[analog obsidian]

or also cap the fps using the nvidia control panel

[analog obsidian]

riivatuner does a better job

try it

reduces stuttering by a lot

[paper bloom]

ohh okay alr so first step imma limmit the fps and second ill install a newer version of this the voice changer=

can i still keep my current one incase i done something wrong?

[hybrid dust]

got a question about smth, so im using ai voices, and when i run it through vb virtual audio cable to discord it echos what people are saying

as in it picks up any output sounds and plays them back through input

i tried different headsets / mics too it still would echo whatever is heard back into the mic

could it be the ai model thats usually the issue or the virtual cable

[analog obsidian]

sure

for gaming try chunks around 150ms-120ms

[low shard]

vb audio cable gives rnadom issues on windows, reported by users there

did you use a youtube tutorial?

[hybrid dust]

yeah! i downloaded it and watched a yt tutorial

[low shard]

rip

video tutorials are outdated asf

[hybrid dust]

but it wasnt like overcomplicated

[low shard]

they use an over year old version of original wokada

[hybrid dust]

i just downloaded the install, hit install drivers and restarted

[low shard]

with outdated settings

[hybrid dust]

ohhhh

any suggestions on what i should do

or like a better virtual cable

[low shard]

it would be better you uninstall what you got off that tutorial

then tell me what's your pc gpu

[hybrid dust]

that wont input everything heard

[hybrid dust]

rtx 4060 is my gpu also i can uninstall the driver

[low shard]

the wokada deiteris fork has better performance and noise suppression along with echo

[low shard]

dw the issue isn't about the drivers

the gpu is good

[hybrid dust]

ill google that and look up a tutorial on how to download it

[low shard]

https://rentry.co/forkvoicechangerguide is the tutorial

[hybrid dust]

ohhh, do i uninstall vb?

[low shard]

I was just going to give u it lol

[hybrid dust]

oh thank you so much!!!!

[low shard]

yes, along with the wokada you got off youtube lol

in the wokada deiteris fork guide, you will get the program, along with a better virtual audio cable

[hybrid dust]

i didnt download wokada!

i just heard about it now

[low shard]

what program did you download? something named mmvcserversio or vcclient?

[hybrid dust]

mmvc!

[low shard]

because that's wokada lol

[hybrid dust]

ohh

it is

[low shard]

we commonly call it wokada

[hybrid dust]

ohh makes sense now thank you for clarifying!!!!

do i uninstall them

and just download the ones through the tutorial u gave me!

[low shard]

yup, let me know for any issues

[hybrid dust]

although i dont exactly know how to uninstall wokada do i just delete the folder?

[low shard]

yeah

and for vb audio cable, you should be able to uninstall it just in the windows app settings

[hybrid dust]

perfect! ill reply if i run into any issues while following the tutorial

[hybrid dust]

perfect thank you so much!!!

[low shard]

You're welcome

[paper bloom]

alright its better then before but still laggin ill try to downlaod the new version tmr ^^

[analog obsidian]

try a higher value than 80

for the chunk

[upper furnace]

sorry one thing can be used to clean up datasets much better?
if little background noise remains
I have to do it for an entry from a cartoon with effects

[upbeat sparrow]

You want to isolate the voice ?

[upper furnace]

yes

[paper bloom]

i used 192

[analog obsidian]

lagging as your voice randomly cutting
or lagging as the game stuttering?

[paper bloom]

voice is like twiching

the game doesnt lagg at all just the voice

[hybrid dust]

quick question for this part! do i need to do this or do i skip it if i have gefore experience

[low shard]

yup you need this, this means it's the nvidia windows version of the program, not that it's the nvidia app

I will go make that clearer in the quick, i feel like you aren't the first confusing that part

[analog obsidian]

i see
well, maybe the fork and vac lite might fix that, but im not 100% about that, marvel rivals is just badly optimized and the 4060 is not a strong gpu

[hybrid dust]

ohhhh makes sense thank you!!!!

[hybrid dust]

it does sound somewhat confusing since it says download nvidia on windows and not the voice changer version for windows!

[paper bloom]

oh okay

[analog obsidian]

i have a 4060 myself
was able to run fortnite dx12, everything low, using the fork with a chunk of 150ms, rmvpe f0

marvel rivals for sure is more demanding than that

even at the lowest possible

[low shard]

edited it

[hybrid dust]

perfect! it helps clarify, and if possible it would help if what was being downloaded was mentioned! not on the title but closer to the bottom something along the lines of "nvidia gpu voice changer download"

thank you so much by the way for the help!!!!

[hybrid dust]

quick question! do i run mmvcserversio once im here or do i run something different in internal

wait nevermind i found it

[upper furnace]

it gives me error if i put voc_fv4 on UVR what do i do?

although I put the files well

[eternal aurora]

@low shard im getting an error running apolo in kaggle

[red obsidian]

when i run start_http.bat the terminal closes after being open for half a second

[low shard]

lemme guess, you're using a youtube tutorial?

[red obsidian]

yup

[low shard]

those are outdated

[red obsidian]

oh

[low shard]

you're using an old version of original wokada

[low shard]

what's your pc gpu?

[red obsidian]

amd radeon 6600 xt

[low shard]

wrong ngrok token

[glacial finch]

worked like a charm like a week ago but now i get this pop up if i open it as usual

[low shard]

wokada has 2 main versions:

• original made by wok
• deiteris fork (modified version) made by deiteris

each with its own updates

we suggest the wokada deiteris fork, which will be way better than the one year old original wokada said in youtube tutorials, read the guide here: https://rentry.co/forkvoicechangerguide

[low shard]

run MMVCserversio

[eternal aurora]

okk

[low shard]

elaborate:

• your pc gpu
• what tutorial link are u using
• what's ur browser

[glacial finch]

4070ti super
the one i got from this server
brave

[eternal aurora]

like this?>

[simple ore]

dont share you tokens bro

I stole it

yoink

now I'm gonna use it

[eternal aurora]

whoops

nooooo

i got this error

[low shard]

have you maybe tried using chrome or firefox ?

[glacial finch]

nope imma do that rq

ye it does work on microsoft edge if i just copy past the url

odd that it doenst work on brave

[low shard]

it might be better you use that, some specific browsers could run into issues

could also be related to some specific settings, but not sure which

[glacial finch]

appreciate the help

thanks!

[low shard]

you're welcome

[eternal aurora]

i fixed it but now im getting this error

@low shard

[low shard]

are you using the applio kaggle from https://github.com/IAHispano/Applio/blob/main/assets/Applio_Kaggle.ipynb ?

[eternal aurora]

yep

you sent me the link and im using that

[static wedge]

Are people still using Mangio RVC for voice conversion? I have an old version and am wondering where it is maintained so I can update. Or if people are using something better nowadays

[odd shale]

Mangio is pretty much outdated and abandoned since 2023.

Unlike Applio which is up to date.

You can either download Applio or use it on colab/kaggle notebook.

[static wedge]

Oh word thanks, good to know. Is it that same git repo linked above?

[odd shale]

Nope.

https://github.com/IAHispano/Applio

This is a link to Applio's repo where you can download it

[eternal aurora]

Is this issue related to proxy settings?

[onyx crag]

why do ppl ear my computer sounds and i can't while using voice changer ? with virtual audio cable?

[brittle wing]

how do i fix the model depending on my gpu

I'll change the model's epochs to somethin' longe no it's overtrained

I need to rework the model

my model said it sounded like a robot

[analog obsidian]

use a good quality dataset and don't overtrain it

[brittle wing]

the model doesn't have any voice inside

[analog obsidian]

then train a normal human voice

[brittle wing]

What's the good epoch for a human voice model

[analog obsidian]

https://docs.aihub.gg/rvc/resources/training/#epochs--overtraining

[brittle wing]

I'm already using tensorboard

[analog obsidian]

you clearly didnt read the part where it explains what is the best epoch amount

[hybrid dust]

im trying to use it but to be honest its really buggy / laggy on the ip thing!

and its using my cpu

also i cant for some reason change the chunk and extra or the usage!!!

[analog obsidian]

stop the conversion

[hybrid dust]

What's that

[analog obsidian]

literally click the stop button

[hybrid dust]

Oh 😭

Sorry, thank you!!

[brittle wing]

What kind of epoch is undertrained

below what

[analog obsidian]

if the voice sounds muddy, different than the voice of the dataset and robotic, it's undertrained

[brittle wing]

what epoch is mainly undertrained below?

[analog obsidian]

bro

train 200 epochs

[brittle wing]

Mine is 300 epochs above not by 400 or

[analog obsidian]

and train a human voice

10 minutes minimum

batch size 4

200 epochs

this is not really optimal, but it'll work regardless

[brittle wing]

I already trained my human voice

My model took about 18 minutes and 52 seconds

to finish

[hybrid dust]

i tried the newer version with like the lite cable but it still echos everything that plays on

any output is played back into input

[brittle wing]

It still sounded robotnik

I checked the model on tensorboard

[analog obsidian]

show settings

[hybrid dust]

i just increased the sensitivity to -66db

[analog obsidian]

f0 rmvpe

[hybrid dust]

o let me change that rq

[analog obsidian]

crepe only for crepe models

[hybrid dust]

how do i know if ones a crepe model

[analog obsidian]

the post should say which f0 was used for training

anyway, crepe is ass for realtime

slow, cant handle noise, huge perf loss

now enable echo in the voice changer

extra 2.7s

chunk 150ms

[analog obsidian]

and sup2

then place your mic to a good distance from your mouth

[hybrid dust]

okay!!! ill turn it on

[rich stone]

Hello, can someone help me create a realistic girl's voice? So, choose a model and adjust it correctly. I would also pay if someone helps me successfully. Just DM me if you can help.

[hybrid dust]

okay let me try this and get back to u!!

[hybrid dust]

it still echos!

anything people say in a vc plays back through my mic

almost like mimicking everything going on thats being outputted right back in

i have the sensitivity at -60db, echo and sup2 on, tried 2-3 different ai models

and 2 different microphones

[knotty moth]

use headphones that doesn't leak to the mic, also turn down the system volume to listen

[analog obsidian]

your headphones might be too loud

[hybrid dust]

i did! i have 2 different headsets

[analog obsidian]

or you have this enabled

[hybrid dust]

onces a razer and ones an okuma, the razer has a bleeding issue every once inawhile but the okuma doesnt, it still does echo

[hybrid dust]

let me check!

i dont believe i do though 1 second

[hybrid dust]

its off!

its at 63 for the mic

ill try lowering my headsets levels

i lowered them it still does echo sounds

[knotty moth]

try even lower

decent headphones shouldnt have issue like that

you can also use any 3rd party noise suppression

if you have some audio routing software other than virtual cable, check the settings

[ancient karma]

Could someone PLEASE!!!!!!!!!!!!!!! give me a RVC trainer that actually works???????????? I don't have a good graphics card, PLEASE

[ancient karma]

Do you know any RVC coach that works like Google Collab?

[brittle wing]

My dataset is quiet

My dataset is big

But it will take too long

what am i goin' to do my model might be overtrained

I have a lot of audio files

@dusty rampart how do I make model not overtrained when there is a lot of audio files on my gpu

[crude flame]

💀

[brittle wing]

I have 22 audio files and how can I make it

[hybrid dust]

20% is what stops echoing 😭 I can barely hear anything

[hybrid dust]

I'll try voicemeeter Banana

[red obsidian]

can someone send video tutorial for the virtual audio cable thing

[hallow thistle]

Don't expect anyone to make a tutorial about Virtual Audio Cable for you. The program itself has nothing much to do than providing virtual Line In device. Also, Virtual Audio Cable lite is a free one. https://software.muzychenko.net/freeware/vac470lite.zip

[knotty moth]

could you double check and show the audio inputs/outputs in voice changer and any applications you're using?

[red obsidian]

mbmb
I got vac lite but I was seeing if anyone could send a YouTube vid since I was confused on the setup since I’m stupid

[knotty moth]

also check this setting for mic and other devices

[torn bane]

where do i download gabox's voc_fv4 for UVR

[brittle wing]

How do i fix the robotic voice in this audio file

But when the output is music it sounds robotic

[hybrid dust]

these are my discord and voice changer settings!

ive tried with echo sup1 and sup2 off, echo and sup2 on

and it still doesnt help with the echo!

[hybrid dust]

check as in turn it on or make sure its off on most / all other devices!?

i checked

all of it are turned off!!!!!

[knotty moth]

the voice changer input should be a real mic, not the one where some audio is forwarded to

[hybrid dust]

its a real mic! its a headset and microsoft recognizes it as mic in at front panel

its realtek

[knotty moth]

the voice changer monitor or discord output is the real headphones ur listening to

[hybrid dust]

wdym!

[knotty moth]

ask chatgpt bro

[hybrid dust]

LMAO

[knotty moth]

the real devices usually display their name, instead of just "Realtek"

[hybrid dust]

😭

LOL

[hybrid dust]

i got the same thing 💔 it does echo still tho

urs shows

the headset tho

idk why mine doesnt

[hybrid dust]

so this would be right

monitor has to be my headphones (output)

[knotty moth]

no idea it may depend on your headphones and the connectivity, but mine is headphones connected through its USB extension

[upper furnace]

Help

it gives me this error if i put voc_fv4 on UVR what do i do?

[analog obsidian]

this is the second time we have someone complaining about echo while using "mic in at front panel" interesting

maybe its the jack connection

some weird interference that doesn't happen with usb

i don't have a jack headset with a microphone to try it
but if this is a real problem, may be good adding it to the guide

[crude bolt]

@hybrid dust , can you go to windows menu > Realtek Audio Console > Device advanced settings and have a look there ? Also what you can see there like is it on an Asus motherboard or whatever rerouting or audio effects are enabled there?

This includes checking the connector settings.

[hybrid dust]

i believe its on an msi motherboard

my pc is msi

[crude bolt]

can you see if the recording device : Tie up same type of input jacks , makes a difference compared to your current 'seperate all input jacks...' option

i expect its not a checkbox but a radio button behavior , by switching the other option the previous will be turned off ... i dont expect them both to be on.

[knotty moth]

try using back panel jacks or the headphone's USB connector like mine

[hybrid dust]

it is a switch option!

[hybrid dust]

let me try that real quick

[crude bolt]

i expect over here those front + back mirrorring could introduce that echoing problem. Especially since they try to split in a software solution. (you can choose) ... the backports are the defaults inputs.

[hybrid dust]

ive switched it to the rear panel! like the mic and headphones jacks

[crude bolt]

the aim is to see if your line1 virtual cable still echos or stops echoing.

[hybrid dust]

and these are the current settings

[hybrid dust]

im about to test it!

[upper furnace]

?

[crude bolt]

also you might recheck weather this option has an influence on windows > sound settings > more sound settings > recordings , alot of devices , check one by one : Properties > Listen > [ ] Listen to device option is off.

you checked that earlier. but to be sure it hasnt sneaky being turned back on.

if after this it still echoes im out of ideas ... maybe the good old ... "eh ... reboot windows" - magictrick works.

[hybrid dust]

ive figured out 1 solution and its lowering discords output to 20% but the con is i can barely / faintly hear anybody :(

[hybrid dust]

let me double checks

check*

[crude bolt]

if the headphone you use have like microphone too close to the speakers then yea. try another headset or a seperate mic

since you use jack ... sometimes that split cable connector thingy is messed up.

where it goes from 1 plug to 2 plugs

and then you experience also in your headphones a not stereo but a mono ish sound

then you have litterallery a short there

So basically you have alot of things to check:

• microphone to close at headset scalps, sound bleeds into mic: lower volume
• cable splitter rec+output messed up: check a seperate mic
• front/backpanel mirror: dont let motherboard see them as seperate devices.
• using front panel: please try back panel
• windows messed up: restart windows
• sound drivers / external 3rd parties listening to device: turn it off
• sound properties listen to device: turn it off.
• you think using a mic but another mic picks up the sound: disable the mics 1 by 1 to eliminate the issue

hm: Okay yesterday i introduced myself a loop accidently by having the sound of my virtual cable feeding straight back into the voicechanger. 😂

but then it echoes nonstop getting louder n louder lol.

[hallow thistle]

[crude bolt]

https://tenor.com/view/galaxy-quest-never-never-give-up-gif-4398729

good luck - i g2g 🙂

[knotty moth]

https://tenor.com/view/cat-cute-cat-yap-yapper-yapping-gif-5642199211123099306

[thorn sapphire]

Hey can have a question where and how to make ai voice model with TITAN model

[formal wind]

Simple answer. Don't figure it out. Figure out how to use normal pretrain

Titan forms noise

[knotty moth]

it's actually more prone to robotic sounds

[lyric fossil]

im new to pretrain models, do all models support universal languages?

[vast elm]

not exactly, some models have a larger dataset in a specific language, in result having better quality when training on that language, you could usually find the training dataset info to find what model is best for the language you want to train your model on

e.g

https://i.postimg.cc/Dy32rfjD/image.png

[untold marten]

Heya, i have a small question. Is there a way/ feature that allows u to like change the "voice"/"process it" just a percentage while using? like 5/10/15/20/40/60/100% (well currently it's working at 100%)

for example:

I wanna show a friend that I "trained" (vocal cord lessons) my voice and can change my voice from original one that i have to the new one gradually.

Think of something like this: a person does an exercise and just spells "a" letter from base voice and gradually changes the tone to higher sounds.

That's a feature that could be fun/made for jokes and while i do have 0 knowledge of how the okada deiteris fork is done, i believe it should be possible but may need to do additional processes like for increments of 5% to go smoothly from original voice to the model that u chose without putting too much stress on gpu or idk.

please tell me if there's something u don't understand