OpenScreen development journal

Software has always stopped at its UI. A video editor is a fixed set of functions: a menu of effects, a list of transitions, a panel of settings. Someone chose them and shipped them. When the thing you have in mind is not among them, you do without.

OpenScreen 2.1 introduces Off-Menu. The built-in functions stay: they are the menu. Next to them sits a new layer: a low-level code interface on the video itself, where an AI agent writes the function you are missing. It has far more freedom than any menu allows, and it still works inside the editor's infrastructure: the timeline, the cursor and the clicks, the zooms and the 3D camera, one renderer for the preview and the export.

Our first implementation is visual effects. You describe an effect, the agent writes it as a GPU shader, OpenScreen compiles it, and it lands on the timeline with its own controls. The preview and the export show the same pixels, on Windows, macOS and Linux.

A ripple on every click. The agent wrote it from one sentence, in two iterations.

v2 started from one sentence rather than a feature list: a demo looks good without touching anything, and touching the settings can't make it ugly. That is the promise the best commercial recorders make, and on September 24 I audited OpenScreen against it.

The result was two rules kept out of ten, and thirty findings. The fixes are gathered in one branch, #814, which is what v2 will be.

Every screen recorder claims a fast export. None of them says fast compared to what.

So the comparison is public now: screen-recorder-benchmark. Same clip, same edit, same machine, across the desktop apps built to turn a recording into a finished demo. OpenScreen is one of the tools in it, alongside Cap, FocuSee, Recordly and Screen Studio, and the standings are whatever the submissions say.

Two things in this stretch matter more than the version numbers. Capture on Windows and macOS now writes fragmented MP4, so a recorder that dies halfway through leaves you a playable file instead of a corrupt one. And the app finally installs the way people on each platform expect it to.

That is v1.10.0, August 24. Before it, v1.9.0 through v1.9.6 in under two weeks.

Exporting a 1080p60 project with full effects used to run at about 8 fps. The same project now exports at about 126 fps on the same laptop. The bottleneck was never the encoder, which is the part of this worth writing down.

v1.8.0 shipped that on August 4, alongside an optional AI editing layer. It transcribes on your own machine, and chat edits go to the LLM provider you connect.

Two releases in the first month, v1.6.0 on July 4 and v1.7.0 on July 19. The second one is the one worth writing about. Eight people who had never touched the project shipped most of it, and several of those changes fix things that had been broken since before the archive.

The last commit on siddharthvaddem/openscreen landed on June 6, 2026. It bumped the Nix package to v1.5.0. Then the repo went read-only, like the README had been warning it would for a while. 39k stars, and v1.5.0 as the final release.

I picked it up on June 15, with the original author's approval. Same name, same MIT license, new URL. This post starts a journal of what happens next.