Github - Blanc-dot
Discord User - https://discord.com/users/824922747423031359

Special thanks to the following people : lusbert, poopmaster, felt, fazemasta, antasma, shadictl, x_hina, sushi

thanks are for anything added to guide, taken from any talks, settings added when previously collecting stable settings, or just apart of other links

---

Guide for W-Okada's RealTimeVoiceChangerClient

---

Other Links

---

Antasma's Local Error Fixes

Antasma's Colab guide

Sushi's useful Links - You need to follow this if you are on an AMD or INTEL ARC graphics card

Frequently Asked Questions

---

Jump To The Important Sections

---

---

Tutorial

---

If you don't have a GPU, either local install the NVIDIA version, or go to the google colab version which will just run better than pure CPU

---

Download NVIDIA

---

---

Download AMD/INTEL GPUs

---

---

Download Mac

---

---

Online Hosted Alternatives

---

---

Virtual Audio Cables (mandatory)

---

• VAC by Muzychenko you should run setup64 after extracting the zip to a new folder not the 64a just the regular 64, you do not need admin for it, and you do not need to restart your computer.
• VB Audio Virtual Cable you should extract the zip to a new folder and run VBCABLE_Setup_x64 as ADMINISTRATOR you will then need to restart your computer

This is what is needed to use the voice changer as an actual microphone input. Otherwise you will be the only one who is able to listen to it.

After installing the Virtual Cable of your choice (VAC LITE {zip} preferred) press WIN+R type 'mmsys.cpl' and change your default audio devices back to your headphones and microphone, make sure to do it for both the regular default and communications default.

---

Updating

---

When updating rename your original folder, download your updated one and try it, if it works, you can copy paste your old model_dir folder to instantly have all your old models back, (every update ive done has worked like that)

if you yolo and want to delete things immediately keep the following

• model_dir folder if you want to keep your models
• your shortcut to start_http.bat this realistically should be somewhere else anyways? but in case it's bugged and has to be in the folder still
• VBS script if you made it.

Just remember to change your extra, chunk, and audio settings back to one value higher than revert as sometimes it'll display the wrong values.

---

Opening on Windows

---

• Unzip your downloaded Zip you acquired from either this guide, or directly from huggingface (it will be the same regardless)
• Grab the main folder called MMVCServerSIO (if you extracted to a new folder) and drag it onto your Desktop.

• Open the folder and find start_http.bat or start_http if you have file extensions hidden and run it,

  • if it instantly closes, click the top bar showing file path, and type cmd
  • if the cmd window doesn't show your folders file path, aka Desktop\MMVCServerSIO you need to type cd Path\to\Desktop\MMVCServerSIO where Path\to\Desktop is your actual file path, you can copy it from the file explorer window by clicking that top bar again and instead of typing, just pressing ctrl+c and pasting it with ctrl+v then you just type start_http.bat in the cmd window to launch the voice changer, this should make the normal double clicking of bat work after it installs everything
• Once start_http is running in your CMD window DO NOT CLOSE IT, let it fully download everything, It will launch the voice changer after everything downloads

---

Opening on Mac

---

• Unzip the downloaded file.

  • Next, run MMVCServerSIO by holding down the control key and clicking on it (or right-click to run it). If a message appears stating that the developer cannot be verified, run it again by holding down the control key and clicking it (or right-click to run it). The terminal will open and the process will finish within a few seconds.
  • Next, execute the startHTTP.command by holding down the control key and clicking on it (or you can also right-click to run it). If a message appears stating that the developer cannot be verified, repeat the process by holding down the control key and clicking on it (or perform a right-click to run it). A terminal will open, and the launch process will begin.
• In other words, the key is to run both MMVCServerSIO and startHTTP.command. Specifically running MMVCServerSIO first.

---

Opening On Multi PC setups

---

On host machine aka the one that will be giving the voice changer to other computers and taking the performance hit

• Open CMD and type ipconfig
• the ipv4 address listed here is what you will need to type on your other machine followed by :18888
• Run the start_https file (or starthttps.command if mac) and go to your other computer / if for whatever reason it doesn't work, you can use the start_http (starthttp.command if mac) file instead
• Open a CHROME based browser, if you want the lightest one possible, grab Ungoogled Chromium,
• in the browser tab type ip:18888, where the ip is what you got from the CMD window before
  EX: 192.168.1.123:18888

---

Uploading Models NVIDIA MAC & Colab/Kaggle

---

• When you have Okada Open, Look for the edit button and press it
  • Pick any slot you want, even if there is already a model there and press upload
  • You do not need to change the model type off of RVC as it supports v1 & v2 and any model you get from AI HUB will be RVC, unless you have a different model type like from SVC or BEATRICE it should always be RVC
  • You do not need to select INDEX at all, in my opinion it is not worth using, but you can also upload it too.

Colab Users do not upload an index file if you are uploading from inside the Web UI as it will just freeze and never let you do so. Just the model.pth is all colab can take in one go. That is your fix to being able to upload models, otherwise follow the tutorial and make sure you use Hina's version as it seems to be working,

Kaggle Users I believe you can just upload both normally, but if it freezes follow whats said for Colab

---

Uploading Models AMD/INTEL GPUs and fixing it's laggy issue

---

• When you have Okada Open, Look for the edit button and press it
  • Pick any slot you want, even if there is already a model there and press upload
  • You do not need to change the model type off of RVC as it supports v1 & v2 and any model you get from AI HUB will be RVC, unless you have a different model type like from SVC or BEATRICE it should always be RVC
  • You do not need to select INDEX at all, in my opinion it is not worth using, but you can also upload it too.
• AMD & ARC users, you need to select the model you just uploaded, and a new button will appear on the UI called export to onnx
  • you need to press this button and go back through the edit menu and overwrite your old model with the new onnx model that was just made from pressing that button.
  • then you need to switch to any other model, and go back to the newly uploaded one for it to register the model type has been changed.
• Reminder that for AMD & ARC after you have your model in ONNX Form, go through each GPU Selector option and whatever isnt laggy like the rest, will be your GPU and that will be what you leave it selected as, alternatively you can look at task manager and check the performance tab as it should tell you exactly what your GPU is number wise.
  • You should be on the recommended settings while testing, open task manager and navigate to the performance tab and note what it says about your GPU

TLDR FIX : Export to Onnx, Reupload the Onnx'd version of model, change voices for files to realize it has changed.

---

Audio Setup

---

S. Threshold: The minimum required sound for it to start converting audio, if you raise IN Gain at the top nothing you do here will matter. If its below it is considered silence, hence the name Silence Threshhold

Voice Changer

INPUT - Your Microphone
OUTPUT - Virtual Audio Cable this is mandatory so you can actually use the voice changer on other apps.
MONITOR - Headphones if you want to have live monitoring, this will throw most people off as there is a delay, but its good for the initial test.

Client mode is easier, and should work IMMEDIATELY.
IF Server Mode whatever Driver (ASIO > WASAPI > MME) you pick they all have to match, you also need to have all your SAMPLE RATES match in MMSYS.CPL

Raising IN gain, raises all the background noise in your mic, get something like SteelSeries Sonar and you can remove all the excess noise before it even hits okada.

Game \ Discord

INPUT - Virtual Audio Cable
OUTPUT - Headphones

---

How to Find Settings for Gaming

---

• Start with the very basic settings for gaming 192 Chunk with 8192 Extra
  • Launch whatever game you want to be playing, and CAP YOUR FPS ideally use RTSS to do so
  • With your FPS capped in game you should change your output in okada to your headphones, so you can hear any audio glitches that will happen when you mess with the chunk value, you will ideally decrease it to as low of a value as you can without it doing the following
    • lowering fps
    • causing the voice to stutter or repeat itself
  • Whatever value you find here should be the chunk value you always use, just get used to the delay it has, and never worry about it again.
• Lastly you can mess with the Extra value, aka something that can increase clarity on your voice, it's not much but keep in mind it does use more CPU usage the higher it goes.

---

Settings (Baseline)

---

FCPE has released, use this, less resources, technically can run lower chunks, but it is slightly worse in "quality" but it's definitely worth using.

---

GPU Chart for Known Working Chunk&Extra on the Official Okada (prebuilt binaries that are linked above)

---

---

---

This is best case scenario, aka good internet speeds. The slower your internet, the higher the chunk value will have to be, for online hosted instances.

• normal users with avg internet, ie under 100mbps should start with 192 chunk, and the same extra values as above

---

Useful

---

Enabling max performance on Nvidia

---

The most important thing of the W-Okada Voice Changer is your GPU clock speeds, in NV Control Panel, enabling "Prefer Maximum Performance" on your card already makes the performance infinitely better. The next thing you can do is undervolting.

Manage 3D settings > Global Settings > Power Management Mode > Prefer Maximum Performance

---

Explanations

---

---

Noise Suppression not working

---

If the inbuilt noise suppression isn't enough just grab yourself Steelseries Sonar, it is the safest method for noise suppression, and it works on every windows setup, with little resource usage.

Mac users the only thing i know of is Krisp, but im pretty sure its a paid thing, if you find a free alternative go to the top, and DM me.

Linux users you have audio effects you can just integrate into your mic using easyeffects or pulseeffects

---

Crackle Fixes (Windows)

---

open task manager hit details

right click audiodg.exe and set Priority to HIGH

right click it again and press set affinity and only have an EVEN number CPU core set, 2 is what I use. AKA everything unchecked except CPU 2

This gets rid of any potential static or crackles from happening.

ProcessLasso can make the audiodg.exe setting continuously reapply every boot

---

Discord Crackle Fixes

---

Input : Virtual Cable (you can rename your virtual cable in the classic windows sound control panel like I did below)

Output : Headphones (or steelseries sonar chat)

Echo Cancellation : Turn this off, when testing in voice & video it causes your voice to cut in and out

Audio Subsystem : Legacy if you are still experiencing crackles, if not Standard.

---

Good models to use

---

Gawr Gura E-Girl Voice (Tea's Voice Model) Minecraft Villager

---

QOL Only for Advanced Users who have ZERO ISSUES.

---

ignore the blue text, just copy all three lines, paste & save as file ending with .vbs