martindelophy/ai-video-editor

Timeline Studio is an AI video editor that runs entirely inside a web browser, similar in feel to apps like CapCut but with a set of AI features built directly into the editing timeline. Because the editing and many of the AI features run locally in the browser, project media generally does not need to be uploaded to a separate server to use most of the tools. The editor gives you a multi track timeline where you can place video clips, picture in picture overlays, captions, stickers, a voiceover track, separated audio, and music, each on its own independent track. You can select clips directly on the canvas to move, resize, rotate, mask, or animate them, apply filters and effects, adjust speed, and set keyframes. It supports snapping, alignment guides, splitting and duplicating clips, zooming the timeline, undo and redo, and saving projects as a portable timeline file you can reopen later. On the AI side, it can generate voiceovers in multiple languages using offline speech models, automatically produce captions from spoken audio using a version of OpenAI's Whisper speech recognition model, detect the main subject in a shot for smart cropping and background removal, separate vocals from music, and even animate a still photo into a talking avatar using audio driven motion and neural rendering. Export runs through a separate, deterministic offline path that composes the final MP4 or WebM video, keeping the preview and the exported file visually consistent. The project also ships an installable agent skill for AI coding assistants such as Claude Code, Codex, GitHub Copilot, and Gemini CLI, which lets those tools inspect a project file, propose specific edits with stable clip identifiers, and apply or render changes through a command line interface rather than only producing an opaque video file. To try it locally, you need Node.js 20 or newer and a modern Chromium based browser, with WebGPU recommended for the heaviest AI features. After cloning the repository and running the install and development commands, the app opens in your browser, downloading AI model files on first use and caching them for later runs. The project is released under the MIT license.