A fixed VTuber avatar reacts over a live cooking demo

This prompt overlays a Live2D-style VTuber commentator onto live-action footage of a Qingjiao Rousi stir-fry, timing its reactions to specific cooking beats. It suits anyone testing how well a video model can hold two independent visual layers, a fixed avatar and a moving camera, without either one drifting apart.

Model: Seedance 2

Use the character from the reference image as a VTuber-style live commentary avatar. Maintain the impression of a green polygonal mask, black borders, narrow eyes, black eyebrows, a large purple tongue, a white head, and black clothes. The character is overlaid as a small 2D/Live2D-style streamer avatar in the bottom right of the screen, not