diffusionstudio/editor

Diffusion Studio is a video editor built for AI agents that converts edits into code and code into video.

The tool addresses the challenge of automating video production by enabling AI agents to perform complex editing tasks through a code-first interface. When you provide raw footage and describe the desired output in natural language, an agent handles the entire workflow—cutting clips, removing filler, adding subtitles, applying color correction, generating motion graphics, and rendering the final video. Every edit made in the interface is written to real code, ensuring the agent always sees the latest version of the project. The tool supports integration with multiple AI models including Claude Code, Codex, and others through an MCP server that registers with your agent automatically.

The tool suits workflows where video production is repetitive or where you want to leverage AI agents for tasks like creating social media clips from long-form content, generating motion graphics and title sequences, or extracting and understanding scenes from video. It works both through a desktop application with a full editing environment and in headless mode for automation. The README does not compare it to other video editors or agent-driven tools.

The project maintains active community engagement through Discord and social channels. Development activity shows consistent iteration on the core editing and agent integration capabilities, with the tool expanding to support video understanding features like summaries and scene search alongside traditional editing operations.