SCAIL 2 Three-person action-driven

This is an AI video generation application based on SCAIL 2, SAM3 three-subject tracking, three-person skeleton motion guidance, and loop continuation.

Users only need to upload a reference image of three people or three subjects, then upload a three-person driving video. The system will automatically recognize the movements of the three subjects in the driving video and enable the three characters or roles in the reference image to follow the video and complete the motion transfer.

Unlike ordinary single-person or dual-person image-to-video processes, this workflow is specifically designed for three-person/three-subject group motion scenarios. It uses SAM3 to simultaneously track three subjects and maintains the left/middle/right character relationship in a left_to_right manner, reducing common issues in three-person videos such as role swapping, missing characters, third-person loss, identity drift, and motion misalignment.

The entire process is automatically completed: driving video reading, unifying the dimensions of the reference image, unifying the dimensions of the driving video, three-subject recognition, three-way mask generation, SCAIL 2 skeleton motion guidance, initial segment generation, loop continuation, overlapping frame processing, color matching, frame sequence merging, and final video output.

The final output is a three-subject motion video, suitable for three-person dances, three-character interactions, three-person motion transfers, group skits, three-character showcases, AI dance videos, group motion guidance for animals or anthropomorphic characters, and packaging into RunningHub video applications.

If packaged as a RunningHub application, the core selling point can be directly expressed as:

Upload a three-person reference image and a three-person driving video to generate an AI video with synchronized three-character motions in one click.