Coding — Python · CLIP · Qwen 3.5 · Flux optique · MCP

Framist

A director's visual reference library: every shot of the best commercial films, indexed and found in a single sentence.

Framist indexes and keeps watch on the best films and photographs, cuts every film into shots and indexes them one by one. You search the way you talk to a DoP: “car at night in the city, dolly in”, “woman turning around, hair in the wind”, “Fermob”, “perfume splash”. The AI understands French and English, film vocabulary (camera moves measured frame by frame, shot sizes, lighting, lenses), objects and brands. Colour galleries, shots as light animated loops, a page for every film with credits and link, projects downloaded in one click. Everything runs on the Mac, in the background, without slowing it down. Internal use: moodboards, decks, ripomatics.

  • Shot-by-shot indexing: every film cut up, every shot described (title, subjects, action, location, lighting, mood) and searchable
  • Seamless search in French and English, with film vocabulary: dolly in, high angle, backlight, FPV drone, anamorphic…
  • Camera moves measured by optical flow: pan, tilt, tracking, zoom, crane, handheld, roll, FPV
  • Brands translated into visible objects: “Fermob” finds metal bistro chairs in every film
  • Colour galleries, clickable palette, tones (warm, cool, dark, vivid…) and shots as animated loops of a few KB
  • Projects: selection, per-project galleries, images and films downloaded as a ZIP with credits
  • Open to AI: documented API and MCP server, so Claude can search and build a moodboard

Screens

Inside Framist.

Home: colour galleries and every image
“car at night in the city”: natural-language search
Film vocabulary: a camera move, measured in every shot
Video mode: every shot as a light animated loop
A film's page: player, credit, download
Nine shots spread across the whole film
Red gallery
“perfume splash”: still life and film vocabulary combined