Getting Started
The current system only has Japanese models for speech processing, so the default contents are in Japanese language only. We actually has an English speech model assets already, and are now working to make English version.
Prerequisites
MMDAgent-EX runs on various platform: iOS, Android, Windows, Mac OS and Linux.
On desktop OS, Audio input device is required. MMDAgent-EX will open the default audio device for audio input.
MMDAgent-EX will use at least 300 MBytes of disk space for system. On desktop OS, all the downloaded files are stored under user’s home or desktop directory.
Network connection is required. MMDAgent-EX will first download the system data from Web.
To make speech recognition work better, avoid using this app in a very noisy environment.
Install
Install the beta app from the download page. Follow the link in the page for your OS, and proceed for installation.
OS | Description |
---|---|
iOS | Tap the app link in the download page and follow the instructions. MMDAgent-EX for iOS is a public beta version, so you may be requested to install TestFlight to intall it. |
Android | Tapping the app link in the download page will open MMDAgent-EX’s Google Play page. Just install it. |
Windows | Download .zip file, unpack it, and put the folder to anywhere you like. |
Mac OS | Download .dmg file, open it, and move the App into Application. |
Linux | Download .tgz file, untar it, and put them anywhere you like |
Launch app
Just launch the app to start! On iOS and Android, tap the installed App icon. On desktop OS, run the executable “MMDAgent-EX” or “MMDAgent-EX.exe”.
At first launch, please wait for a minute until all the system data and default dialogue content has been downloaded from Web server.
Try it out!
After successfull start up, the default dialogue content will start playing. You can talk to the default sample agent. Take a brief look at the following pages to know basics.

Startup screen
Play other content
As an extra, you can try another Web content we have prepared. This content is provided to offer an experience of Web-based content delivering. It is a animation character dancing content, a fun-made one originally built on MikuMikuDance by creators. No voice interaction is used in the content, but you will know that MMDAgent-EX has rich character expression rendering engine on real-time with high frame rate.
Warning
This extra content will show a Japanese 2D-like anime character dancing with anime songs.This content is distributed as a Web content. On iOS, Android and MacOS, just tap the link below to play. On Windows, paste the follwing URL link to the MMDAgent-EX window. On Linux, give the URL as a command line argument.
- iMarine Project - Welcome to DEEP BLUE TOWN! (modified for MMDAgent-EX)
OK, what shall I do next?
Go to the Key Bindings to learn how to interact with agents!
Or if you are developer or content creator…
- Go through System Overview page and Content Overview to know the details of this system.
- All the scene are manipulated by a dialogue scenario. You can learn how to write or edit a scenario. More details of the file formats are described in File Formats section.
Feedback
Was this page helpful?
Glad to hear it! Please tell us how we can improve.
Sorry to hear that. Please tell us how we can improve.