OpenFilm gives coding agents a local, editable workflow for making videos
OpenFilm is an MIT-licensed open-source project that connects coding agents to video creation and editing. The agent works with a project folder containing film.html, an HTML file that composes web pages, video clips, images and audio on a timeline. Studio, a local editor that runs in the browser, lets people change text, layers, colors, placement and clips and saves those edits in the project files. According to the repository, it can export video, GIFs, audio, subtitles, stills and formats for editing tools such as Final Cut Pro, Premiere Pro and DaVinci Resolve. Version 0.2.5, published on October 21, focuses mainly on export: its changelog says editing-project export is two to four times faster and uses less free disk space, and also lists DeepSeek Harness support and Windows safety improvements. The CLI installation requires Node 22.2 or later, ffmpeg and git, and a desktop app is available for macOS and Windows. Making and editing films does not require an account, but generating voice, music, images and video requires connecting services with your own API keys.
Otomatzia editorial · Analysis published
Source published: · OpenFilm
What changed and what is actually open
OpenFilm offers an agent a simple contract: a local folder, film.html and HTML pages where frame(t) draws the picture at each moment. The agent can use CSS, SVG, Canvas, WebGL, video and other web technologies, while Studio presents the same work for manual editing. The repository says the format, CLI, Studio and desktop app are MIT-licensed. These are capabilities and design claims from the project, not an independent test by us.
Why automation builders may care
The folder-and-files model fits workflows already managed by coding agents: a film can live in Git, an agent can be asked to change a page or the timeline, the result can be checked and an earlier version can be restored. Separating video creation from external media services may help teams build a repeatable process. The claim that an agent can make a film does not, however, guarantee production-ready quality, consistency or render time.
The cost of local execution and the limits to check
Before adopting it, check the computer, storage and ability to run Node, ffmpeg and headless Chromium. The README says the first run downloads Chromium at about 100 MB, and that generating voice, music, images and video depends on your connected services and keys. Define which keys the agent may use, where media files are stored and what happens when an external service is unavailable. The stated support, macOS, Windows and Linux for the CLI and macOS and Windows for the desktop app, should also be checked against the team's environment and formats.
A pilot checklist
For a small pilot, keep a short film in a repository, run the same project more than once and verify that frame(t) returns the same picture when frames are requested in a different order. Test the export for the destination the team actually uses, along with disk consumption and behavior on the relevant operating system. Only then connect media-service keys, define permissions and secret storage, and decide which part of editing remains subject to human review.
