Blendshape

A 3D animation technique that defines specific facial configurations (such as a smile, raised eyebrow, or open mouth) as target shapes, which can be blended together at varying intensities to create complex facial expressions. In sign language avatar systems, blendshape coefficients captured from face tracking (e.g., MediaPipe's 52 blendshape outputs) are scaled and applied to the avatar's facial rig, enabling expressive non-manual signs essential for grammatically correct sign language production.