future app · harnesses Letterboxd idea AI-enhanced
encore
Your Letterboxd 5-star reviews, distilled to the one scene that wrecked you. encore stitches the timestamp + your one-line review into a 20-sec kliplet.
"The scene, not the film."
The hook
Letterboxd power-users have already done the work of writing reviews. encore turns that latent corpus into shareable kliplets — by scene, not by film.
The klip-feed
| Replaces | The "this scene though" tweet |
| Klip payload | Poster + scene timestamp + review excerpt + Letterboxd URL |
| Profile signal | A tiled film grid of "movies that hit me" — discovery ordered by resonance, not release date |
AI layer
| What it does | An LLM scans your review prose for emotional-intensity phrases ("wrecked me", "couldn't breathe"), extracts the most quotable line, and infers which scene/timestamp it points to using the film's plot wiki. |
| Input | Your full review text + film's scene-level synopsis |
| Output | Pull-quote + likely scene timestamp + tone tag |
| Why with > without | Without AI, you'd manually tag the scene + line. With AI, your existing reviews become kliplets without you lifting a finger. |
Open questions
- Letterboxd API — public API is read-only and rate-limited; need OAuth for personal review access. Confirm scope.
- Scene clip — fair-use concern. Maybe poster + your text only (no actual footage) for v1. The kliplet sells the quote, not the scene.
- Scene-inference accuracy — LLM may hallucinate timestamps. Show as "approximate" or let users confirm.
Status
idea Filed 2026-05-15 from brainstorm session.
Lino Labs · future-apps · encore