AirDeck

Named

AirDeck. A gesture-driven slideshow: a camera watches the presenter's hand and turns what it does into slide commands, with the recording tied back to the slides.

Problem defined

Presenting is still tied to a button. Clickers die, get lost, and keep you near the laptop. A physical laser dot never shows on a screen share or a recording. And almost nobody reviews their own talk, because scrubbing through footage to find "the part where I rushed slide 4" is a chore, so the feedback loop never closes.

Evidence gathered

To verify with real cases:

• How often remote and hybrid talks lose the laser pointer (it only exists in the room).
• Whether speakers who rehearse on video actually rewatch, and what stops them.
• What existing gesture control (smart TVs, game consoles, presenter apps) got wrong: false triggers, arm fatigue, lighting.

Flow mapped

1. Camera: the laptop webcam, 30 frames a second.
2. Hand model: MediaPipe finds 21 points on the hand, entirely in the browser. No frame is uploaded.
3. Gesture: the points become a pose (thumb out sideways, thumbs up, point, OK sign, open palm, fist, peace sign). Every command is a pose held for a moment, so moving the hand around never fires anything by accident.
4. Command: thumb out to the right or left changes slide, pointing shows a laser (a button or the L key hides it), an OK sign (thumb and index touching, other fingers up) draws ink on that slide, thumbs up reveals the next build, a held palm wipes ink, a held fist pauses the screen, a held peace sign starts or stops recording.
5. Log: every command is stamped with the time and the slide.
6. Review: the log splits the recording into slide segments, so the video has a clickable slide timeline plus time per slide, gestures used, and notes on where the pace dragged or rushed.

Phase 1 deliverable

A working browser prototype on the site: AirDeck demo. A seven-slide deck about the idea itself, all seven gestures, per-slide ink, animated builds, recording with the review screen, and keyboard and mouse fallbacks for when there is no camera.

Refining

What changes once people try it:

• Tune hold times and pinch distance against real rooms, lighting and distances.
• Done: drawing moved from a plain pinch to an OK sign, because the hand model kept confusing a pinch with pointing. The two now differ by how many fingers are up.
• Done: the touch area follows the presenter and can be resized. Swipe was replaced by thumb-out, because a moving hand was too easy to misread.
• Load any deck (PDF or Google Slides) instead of the built-in one.
• Burn the laser and ink into the recording, not just the camera feed.
• Voice commands alongside gestures ("next", "go to slide five").

Action steps

What gets it in front of people:

• Test with three presenters and log every false trigger.
• Add deck import.
• Record a short demo video.
• Done: portfolio tile is live.