Video production as code

Most product videos are made once, in a timeline editor, and then quietly rot. The product moves on; the video does not. Re-recording is expensive enough that teams just live with the drift. We think that is backwards — a product video should be as easy to keep current as the product itself.

So decodx makes a video a committed spec. Scenes, narration, captures, and branding live in plain files next to your source. Rendering is a pure function of that spec: the same input always produces the same final.mp4, plus captions, chapters, and a thumbnail.

What changes when a video is code

• Diffs and review. A wording fix is a one-line change your team can review in a pull request, not a re-export nobody can inspect.
• Cheap iteration. Expensive assets are cached by content hash, so changing one scene re-renders only that scene. Re-renders are free.
• Automation. Because every command is non-interactive, a CI job — or an AI agent — can produce the video without a human at a keyboard.

The result is a video you treat like any other build artifact: authored as source, rebuilt on demand, always in sync with what you ship.