AI modes

Every recording is first reconstructed into a timeline (what app or site, what activity, what changed, what went wrong — in 3–30 second segments). The mode then decides what to write from it. You can re-analyze any recording in another mode later; earlier outputs are kept.

Explain

A 2–4 paragraph narrative of what happened, a list of key moments, "what to know" bullets, and a glossary of unfamiliar UI. Best for handing a session to someone who wasn't there.

Bug report

Title, severity guess, environment (browser, OS, viewport), numbered repro steps, expected vs actual, the first bad moment, evidence lines from console/network/visual cues, a suggested component, and a ready-to-paste issue body. Free plans include Bug Report because it's the wedge we care most about.

Tutorial

Prerequisites, ordered imperative steps each tied to a timestamp and click target, and the end state. Exports to PDF/DOCX with the matching screenshot under each step — an SOP from a single recording.

Recap

Apps seen, topics, decisions and action items (with an owner guess) from a working session, review or call.

Evidence-linked, always

Every item the AI produces carries a timestamp inside the recording, and the renderer turns it into a seek link. If a claim isn't anchored to a moment, it doesn't ship. Thumbs up/down on any section feeds our prompt evaluation.

Silent recordings

Narration is optional. The model reads sampled keyframes, the interaction event log (clicks, navigations, tab switches, console errors) and page-context snapshots. If you did speak, the transcript is used too — and shown in the Transcript tab.

Voiceover

Pro plans can generate an AI-narrated version of any Explain or Tutorial recording: a script timed to chapters, spoken in one of three voices, muxed into a second video variant you can toggle in the player.