SCAIL-2 Dual Character Swap

This is an AI video generation application based on SCAIL 2 SAM3 dual-subject tracking, dual-skeleton guidance, dual-character replacement, and long-video loop continuation.

Users only need to upload a dual-person reference image and a dual-person driving video. The system will automatically recognize the actions of the two characters in the driving video and replace them with the two characters in the reference image at the corresponding positions in the driving video.

Unlike ordinary single-person replacements or image-to-video generation, this process is specifically designed for dual-person videos. It uses SAM3 to simultaneously track two subjects and maintains the left-to-right correspondence of identities, reducing common issues in dual-person videos such as replacing only one person, the second person being lost, left-right identity swaps, character drifting, and action misalignment.

The entire process is completed automatically: driving video reading, reference image size unification, driving video size unification, dual-subject recognition, dual-path mask generation, SCAIL 2 dual-character replacement, initial segment generation, loop continuation, overlapping frame processing, color matching, frame sequence merging, original video audio integration, and final video output.

The final output is a long dual-character replacement video that inherits the actions, rhythm, and camera structure of the driving video while striving to maintain the identity, appearance, clothing, and left-right relationship stability of the two characters in the reference image.

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

Upload a dual-person reference image and a dual-person driving video to generate a long dual-character replacement video with one click.