Cutroom is a native macOS video editor with an AI agent inside. Cut on a real multi-track timeline the way you always have — or describe what you want, and the agent finds the shots, writes the captions, scores the edit, and checks its own work.
 Edit by asking — the agent plans, executes tool calls against the timeline, and streams its reasoning as it works. |
 It checks its own work — after editing, the agent renders a frame of its own timeline and looks at it before calling the job done. |
 Find any shot by describing it — on-device visual search embeds your frames and your words in the same space; search is just distance. |
 Captions, on device — speech transcription with sentence-aware phrasing, styled caption clips, one undo for the whole batch. |
 A score, composed to fit — mark a range and generate music that follows the edit, or describe the track you want. |
 Shots you never filmed — generate images and video from text or references, then bring stills to life. |
Underneath: multi-track ripple editing, linked A/V, keyframe animation, source/program preview, in/out range tools, Final Cut Pro XML export, and self-contained project bundles.
git clone https://github.com/kuluruvineeth/cutroom.git && cd cutroom
swift build && swift runThe AI features are powered by a Convex backend (auth, credits, generation, agent). Full step-by-step setup — Convex, Clerk, fal, Stripe: docs/SETUP.md.