wassermanproductions/motion-previs-studio

What it solves

Motion Previs Studio is designed for filmmakers who need precise control over AI-generated video. Instead of relying on random generation, it allows users to extract specific motion, pose, and camera data from a reference video to create a "production pack" that guides downstream AI video tools (like Runway, Kling, or ComfyUI) to maintain consistent movement and blocking.

How it works

The app processes source video (local files or YouTube URLs) and analyzes it to extract several control layers. It uses MediaPipe Pose Landmarker for body tracking and Lucas-Kanade optical flow with RANSAC similarity fit to solve for camera movement independently of the subject. The tool then encodes these as deterministic videos and JSON data, which can be exported as a bundle for use in AI pipelines or imported into Blender.

Who it’s for

It is built for AI filmmakers and previsualization artists who want to move beyond simple prompting and use reference-based control for actor motion, camera paths, and scene depth.

Highlights

• Comprehensive Control Layers: Extracts OpenPose BODY_25 skeletons, depth maps, edges, masks, and camera motion data.
• Subject-Independent Camera Solve: Recovers background camera movement (pan/tilt/zoom/roll) while masking out the tracked subject.
• Agentic Control: Includes an MCP bridge allowing AI agents (like Claude Code) to drive the app's analysis and export workflow.
• Production-Ready Exports: Generates a "Production Pack" including JSON manifests, Python scripts for Blender import, and formatted videos for ControlNet graphs.
• Cross-Platform: Available for macOS and Windows 11.