D3voz/openjutsu

What it solves

Openjutsu allows you to replace people or objects in existing videos while keeping the original camera movement, cuts, and soundtrack. It solves the problem of needing complex professional editing software to swap characters, change clothing, or modify backgrounds in video footage.

How it works

It acts as a user-friendly interface for ComfyUI, using a combination of advanced AI models to process video. It uses SAM 3.1 to track people and objects, MiniMax H3 to generate the new video content, and Qwen3-VL to describe and understand the visual elements. You simply upload a video and a character photo, and the tool handles the tracking, masking, and frame-by-frame replacement locally on your GPU.

Who it’s for

This tool is for users with an NVIDIA GPU who want to perform creative video editing—such as character swapping, object removal, or relighting—using open-source AI models on their own hardware without relying on cloud-based subscriptions.

Highlights

• Character Swapping: Replace the main person in a video with a character from a photo while preserving their original motion and pose.
• Object Manipulation: Easily remove, change, or add objects like clothing, hair, or background elements by painting over them.
• Advanced Motion Transfer: Supports multi-character scenes where different people can be swapped simultaneously using role-based tracking.
• Post-Processing Tools: Includes built-in features to fix facial details and adjust color grading for a more polished final result.
• Local Execution: Runs entirely on your own machine, ensuring privacy and control over your data and compute resources.