Suggest zooms for a screen recording

Turn the pointer moves and clicks noted beside a screen recording into zoom suggestions (where the pointer settled, for how long, how strong), as plain rows the person can edit before the video is exported

Give sidecar: { space: 'normalized-source', samples: [{ t (ms), type: move | click, x, y (0 to 1 of the captured source) }] }. Telemetry in another space or without one is refused. Answers zooms: [{ start_ms, end_ms, x, y, strength (light, medium, strong), zoom, clicks }], sorted, none overlapping, the same every time for the same samples. Thresholds: read-data/auto-zoom.json.

What it asks for

Nothing: your AI starts it with your words.

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.

Say to your AI: “Use Trillion to turn the pointer moves and clicks noted beside a screen recording into zoom suggestions (where the pointer settled, for how long, how strong), as plain rows the person can edit before the video is exported”