filmprint ingests your full rating history and watchlist — typically thousands of data points — and builds a multi-dimensional taste vector across genre, tone, era, and subgenre axes.

An ONNX embedding model encodes each film into the same vector space as your taste profile. Films are ranked by cosine similarity, so the candidates closest to your actual preferences surface first.

Tell filmprint what you're in the mood for — genre, vibe, runtime. It re-ranks candidates against your mood-adjusted taste vector and uses Claude to explain exactly why each pick fits you right now.