SOMA FORGE
Paint live video onto a MediaPipe-driven body, then export it as a textured, animated, part-separated GLB.
idle
models
source
take
Video in
📷
Camera
🎬
Video file
🖼️
Still image
🖥️
Screen
Looking for cameras…
Flip
camera
— fps
Tracking
●
Body — 33 points
0
●
Face — 478 points
0
●
Hands — 21 × 2
0
Start tracking
Depth exaggeration
1.00×
Smoothing
0.35
Colormap from video
Every frame is copied to a texture atlas. Each vertex reads the pixel it sits on, so the model wears whatever the camera sees.
As shot
Punchy
Black & white
Neon shift
Warm skin
512²
1024²
2048²
Freeze texture
Grab now
Live · waiting for video
Activity
Ready.
building the forge
Skinned
Wireframe
X-ray
Point cloud
Neon
Reset view
Grid
—
render fps
—
infer ms
0
vertices
0
triangles
0
parts on
0.0s
take length
Take & export
Record take
Clear take
A take stores joint motion as keyframes and face/hand shapes as morph targets, so the GLB opens in Blender with animation already on it.
Export GLB
Export each part
Export T-pose only
Save texture PNG
no export yet
Parts
Select all
Select none
Isolate face
Video out
Record 3D view
Record screen
3D view
screen
codec
Saved files
Refresh
Delete all
Nothing saved yet. Exports and recordings land here.
Telemetry