Inspiration As a founder, I’ve recorded countless YC-style founder videos and marketing clips. Every “final” take meant another retake — stopping the camera, walking over to the phone, losing eye contact, then editing out the mess later.
When I saw iPhone Duo, that problem flipped: two screens, a great rear camera, and room for a real director on-device. That’s why we built Director — so creators (and founders) can tent the phone, read the script, get live notes, and pause with an open palm instead of ruining the take. This form factor doesn’t just help the product. It changes how we ship our own story.
What it does Open Director → load a script → tent the foldable → the rear camera rolls.
On the inner display: a live preview, a scrolling teleprompter, and short AI coach notes while you talk. Raise an open palm — recording pauses (take saved). Same palm again — resume on the next take. No walk-over. No touch required.
Optional: hand controls to a second person on the other face (Operator screen). Optional: stitch takes into one polished cut — then post.
In one line: a pocket AI director for foldables — rear cam, live coach, gesture pause.
How we built it We built a native SwiftUI app for iPhone Duo, using ArrangementView for Talent | Operator, reserved regions around the hinge, and an outer subject-facing accessory where available.
Capture: rear-camera-first session with Start / Pause / Resume as discrete takes ML: on-device Vision / hand-pose — open palm (~0.5s hold) toggles pause ↔ resume, with cooldown Studio UX: teleprompter + lightweight coach tips; dual-pane two-person mode Assemble: stitch multiple takes into one cut Demo resilience: Demo Media + Simulate palm so we could finish under Simulator limits Stack: SwiftUI · AVFoundation · Vision · Duo layout APIs · on-device ML.
Challenges we ran into Camera permissions & capture — Simulator has no trustworthy rear camera; Continuity / device permissions needed careful Live + Demo paths. Gesture reliability — confidence, hold time, and cooldown; on-screen Pause/Resume share the same handlers as a fallback. Foldable layout under time pressure — Talent vs Operator split, hinge clearance, state that survives folding. Stitch vs shoot — lock the record → palm → resume loop first; assemble as polish, not a blocker.
Accomplishments that we're proud of Full Director recording loop for the hackathon demo Palm pause / palm resume wired to the same take model as the buttons Dual-screen story: talent script + coach, operator transport Optional stitch for multi-take sessions A demo we can actually show not just slides
What we learned Foldables aren’t bigger phones screen role (Talent vs Operator vs outer) matters more than scaling UI. Creators need fewer interruptions mid-take; gesture pause is the aha. On-device ML works for gesture control with hold + cooldown + tap fallback. Demo infrastructure (Demo Media, Simulate palm) is real product work under Simulator limits. Founder video pain is universal we lived the problem we built for.
What's next for AI Director Polish Live Continuity / device camera and Duo HIG layout for shipping hardware Tighten gesture accuracy and coach quality on real takes Ship stitch + export presets creators actually use Publish before iPhone Duo launch on 26 October
Then: App Store, creator feedback, and the Pro coach / assemble lane
Log in or sign up for Devpost to join the conversation.