Posts in Videos
How an agent's browser run becomes a finished video: recording the page cannot notice, camera and cursor, voice, subtitles, GIFs and the label for an AI voice.
An agent that already operates a product in a browser can film what it does, and the film can be rendered again whenever the product changes. The articles here follow that path from the recording to the file: whether a page can tell it is being recorded, how the camera follows each click, which voice and which subtitles fit a demo, how a GIF stays under a launch page's limit, and which label an AI voice carries. Every measurement names its date and its number of runs.
How the recorder, the renderer and the presets work is described on the videos page of the documentation.