← Back to Pose Extraction overview
PROBLEM: Animations need expressive face/body params, but I have no animator on staff — only video reference of myself recording each gesture.
WHY IT MATTERS: Per-frame face param extraction is the bridge between human-recorded video and Rive runtime. Without it, animations would be hand-keyed (slow) or generic (off-brand).
STACK: Python (RTMLib for 2D landmarks, MediaPipe-style pose), JSON manifest, Rive runtime (Kotlin Android)

🎬 Frame Browser: encouraging

Frame 0/49

Arm L: —°

Arm R: —°

Hip Y: —

Tilt: —°

Motion: —

Motion over time

Arm angles over time

Hip Y over time

rest candidate = low motion + neutral pose gesture peak = highest deviation from rest