💡 Login to apply to projects
Objective: Natural transformation of multiple people in challenge/dance videos by 1:1 mapping to desired persons (or avatars). Each person maintains original dance movements (pose/timing/rhythm) while appearing with the designated person's face/style (hair, accessories, outfit tone). Natural handling of multi-interaction scenes including occlusion, crossing paths, and intersecting movements. Input/Output Input: - Challenge video - One or more frontal images of target persons/characters - Mapping table: e.g., Original Person 1 → Target A, Person 2 → Target B, etc. Output: - Output video (mp4), minimum 5 seconds - (If selected) Production method guide required Quality Requirements: - Movement preservation: Maintain original pose/timing/rhythm - Facial consistency: Maintain consistent appearance throughout video - Temporal stability: Minimize flickering and boundary tremors - Occlusion handling: Maintain consistent layer/depth, clean composite boundaries - Composition/correction: Natural integration of colors, brightness, and background - Standardization: Reliable performance across different input videos and subjects Current Development Stage: Currently developing single-person transformation using ComfyUI + WAN 2.2 model Reference workflow provided Note: ComfyUI not mandatory; other commercial/alternative tools acceptable Deliverables: - Input video and reference persons/characters (original preserved) - Output video (mp4), minimum 5 seconds - Three different video and person combinations - Production method guide required if selected
