Openjutsu Open-Source Tool Enables One-Click Video Character and Outfit Swaps

Openjutsu, an open-source video character and object swap tool, was released on GitHub under an MIT license. The tool accepts a video and character reference photo, regenerates the main person while preserving camera movement, cuts, motion and soundtrack, and supports 1 to 4 characters in multi-person mode. It runs locally on ComfyUI and MiniMax H3, with SAM 3.1 for tracking and Qwen3-VL 8B for image description.
Event version 1.
Key insights
Preserves original video elements
Openjutsu replaces the main person in a video with a supplied character while keeping the camera, cuts, motion and soundtrack unchanged. Users can also change hair, clothes, props or background separately.
Multi-person mode supports up to four characters
The tool supports 1 to 4 characters in multi-person mode, allowing each reference image to be assigned an identity. Characters not selected remain unchanged.
Local execution on open weights
Openjutsu runs on the user's own GPU using MiniMax H3 for video, SAM 3.1 to track people, and Qwen3-VL 8B to describe pictures and shots. The app is a Next.js front end over a ComfyUI workflow.
Preview and segment selection features
A preview marks areas to be redrawn with colors, and users can drag a progress bar to inspect frames. Users can select only the beginning, middle or end segment of a video for a short test.
Additional face repair and relighting
The tool includes face repair and color grading functions, plus 17 relighting presets.