Preview
No signalPreview stage — load a video to mirror frames onto the 2D canvas.
Playback
Native decoding. Volume maps straight to the video element.
Face tracking
On-device MediaPipe landmarks. No upload, no identity — shape only.
Face tracking idle.
⟳ Rotate your phone for the full side-by-side editor.
Mask library
Your masks on live landmarks. Upload a transparent PNG, WebP or GIF to begin — on-device decoration, no upload, no identity.
Start face tracking, then pick a mask.
Upload mask image
Transparent PNG, WebP or GIF. Static images track the face now; smooth animation follow is built in.
Mask actions
Mask adjustments
Fine-tune the selected 2D mask.
3D masks
Your custom GLB models on live landmarks. 2D masks hide while 3D is on — the 2D engine stays as fallback.
3D masks idle.
Upload GLB model
Self-contained .glb only (v1 does not accept .gltf + separate textures).
Source video
Drop a file or browse. Decoding stays on the native video element so audio keeps playing normally.
Export
Frame PNG captures the raw frame. Video export records the masked composite (video + masks, no debug lines) in realtime.
Self tests
Pure parser + utility tests. Same suites run under npm test. Developer mode only.
Integration slots · reserved
Stage 0 wires nothing here. Later stages register without touching playback code.