VTuber Stage — Model No Longer Semi transparent
The model in the VTuber Stage was rendering with a weird ghost/transparent look when placed over a background. Fixed.
The Stage captures the main app window and composites it on top of your background. It was using a blending trick to remove the dark background, but that trick also ate into the dark parts of the model itself — shadows, clothing, hair. The result was a semi-transparent character that looked off on stream.
The fix swaps to a proper alpha channel approach. The model now renders as a solid, fully opaque character the way it's supposed to look.
No settings to change. Just open the Stage and it works correctly.
Flat Avatar — 2D Mode
You can now use a 2D sprite avatar instead of (or alongside) the 3D VRM model.
The idea is simple: you give the app a folder of PNG files — one for the body, one for the face, one for each eye state, brows, mouth positions, and hair — and it composites them into a layered 2D character on screen. No Live2D, no special software, no licence fees. Any art app that can export transparent PNGs works. Krita, GIMP, Aseprite, Photoshop — anything.
A default avatar pack is included so you can try it immediately without making any art.
To load it: Settings → Flat Avatar (2D Mode) → select the pack from the dropdown → Load Model.
What it responds to:
Face tracking — if your camera is on, the avatar tilts its head with you, blinks when you blink, and opens its mouth when you speak. Turn the camera off and it still breathes (slow idle bob) and blinks on its own every few seconds.
AI emotions — when the AI responds with a happy, sad, angry, or surprised tone, the eyebrows swap to match. Each emotion response also triggers a brief scale pulse so the avatar visibly reacts.
TTS lip sync — when the AI is speaking through TTS, the mouth sprites step through open/closed states in sync with the audio amplitude. This works independently from face tracking so both can be active at the same time without fighting each other.
Moving and resizing:
Drag anywhere on the avatar to reposition it. Scroll the mouse wheel over it to make it bigger or smaller. Position and size are saved and restored automatically when you restart the app.
Visibility controls:
The Enable 2D Avatar toggle in the settings section shows and hides the avatar without unloading it — so switching back to 3D-only is instant. The Show 3D model alongside toggle lets you run both the 2D and 3D avatars on screen at the same time if you want.
Making your own pack:
The format is a folder with an avatar.json manifest and a set of PNG files named by convention (eye-left-open.png, mouth-closed.png, etc.). All images need to be the same canvas size with a transparent background.
User guides:
Plain text guides for all the major features are included with the install. You can find them inside the app folder at renderer/assets/guides/ — open any of them in a text editor or file manager. There is one for the 2D avatar format and pack creation, one for adding custom animations, one for face tracking setup, one for hands-free voice mode, and one for OBS integration. README.txt in that folder lists what each file covers.