Export it
Choose Export for AI on the card that appears when you stop recording, or in the export sheet beside Export. It uses its own settings, so the format and quality pickers stay set for your regular exports.
The video lands in ~/Movies/Kirocast/Exports with its .events.json beside it.
Add your words
When you narrated the recording, a transcript lets your agent read what you said as well as what happened on screen.
- From the card, Kirocast transcribes first, then renders, and shows both on the card.
- From the export sheet, Kirocast asks once: choose Transcribe and Export to include it.
Transcription runs on your Mac. The first time, it downloads the speech model once, about 600 MB.
Copy the prompt
When the export finishes, choose Copy Prompt in the sheet (Copy prompt on the card) and paste it into your agent. It lists the files that explain the video first, then the video:
I recorded my screen with Kirocast to show you something. Use the events.json to understand what happened in the video.
Events: ~/Movies/Kirocast/Exports/Export_1787674297.events.json
Video: ~/Movies/Kirocast/Exports/Export_1787674297.mp4
<describe here what you want the reader to do>Replace the last line with your request. A Transcript: line appears too when the recording has one.
What the video contains
A model samples a video into stills, about one per second, at around 1500 pixels on the long edge. The export is built for exactly that:
| Setting | Value |
|---|---|
| Resolution | Up to 1536 px on the long edge, never upscaled |
| Frame rate | 5 fps |
| Format | MP4, H.264, no audio track (your words are in the transcript) |
| Kept | Click pings (with Click animations on), Smart Annotations, Tap Tap cues and privacy masks |
| Left out | Zoom, backdrop, padding, shadow, rounded corners, captions and the camera bubble |
A one-minute capture of a large Retina display exports in about ten seconds to roughly 5 MB, with code and small labels still sharp.