A film stills database you can search by what's in the frame

What's in it

5,489

films

923

director pages

882

cinematographer pages

28

technique pages

5,277

measured colour palettes

464

measured film lists

Per-film frame counts are shown on each film page. Features, documentaries and animation, from the silent era to this year.

What you can search on

• —Lighting. Key and quality — high-key, low-key, hard, soft, practical-lit, silhouetted, backlit.
• —Lens character. Anamorphic, spherical, wide, long, vintage-soft — the character of the glass, not just the focal length.
• —Shot size and framing. Close-up through extreme wide, plus camera angle and how the frame is composed.
• —Colour. Palettes extracted per frame as hex, so you can search a look by its actual colours rather than by a word for them.
• —Mood and era. What the frame feels like, and what period it depicts — which is often not the year it was made.
• —Plain language. Or skip the taxonomy: "silhouette against a window, low-key", "wide anamorphic desert at golden hour".
• —A reference image. Drop in any picture and get frames that match its light, colour and composition.

Reachable by machines, not just by browser

A stills database is only as useful as the things that can reach it. FrameThrower is the reference app for the future creative: the library is callable from a REST API, an MCP server and an npm SDK, so an agent can pull references into a moodboard, a lookbook, a shot board or a production pipeline without a human in the loop.

npm install framethrower-ai

curl -X POST https://framethrower.ai/api/v1/search \
  -H "Authorization: Bearer ft_YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query": "neon-lit rainy street at night", "limit": 5}'

Free tier included. $2 of credits on signup, no card, and they work through the API from the first call — no plan to buy first. After that it is usage-based: $1 = 1,000 credits, a search costs 2, so $1 buys 500 searches. Credits never expire.

Built by filmmakers and ads creatives, for film professionals. People who have shot, cut and pitched for a living — which is why the index is organised around lighting, lens and framing rather than keywords.

API documentation · machine-readable spec · npm package

Who builds it

FrameThrower is built by senior filmmakers and ads creatives — people who have shot, cut and pitched for a living — for working film professionals. That is why the index is organised around lighting, lens and framing rather than keywords: it is the vocabulary the work is actually briefed in.

Rights

The frames are not ours. They are shown at reduced resolution as references for commentary, education and study, and rights remain with their respective owners — we do not grant a licence to them. The measurements we compute from them — colour palettes, counts, rankings — are ours, and those we publish under CC BY 4.0 with attribution. Full detail on intended use.

Common questions

What is a film stills database?

A searchable collection of frames taken from finished films, used as visual reference — by directors and cinematographers planning shots, by production designers and colourists, and increasingly by software that needs real cinematography examples. What separates one from another is not how many images it holds but what you can search on: a gallery lets you browse by title, a database lets you search by what is inside the frame.

How many films does FrameThrower cover?

FrameThrower indexes 5,489 films — features, documentaries and animation — with 923 director pages and 882 cinematographer pages built from the same catalogue. Per-film frame counts are shown on each film page.

What can you search on?

Lighting key and quality, lens character, shot size and framing, camera angle, colour palette in hex, mood, era and time of day — alone or combined, described in plain language. You can also search from a reference image, or by colour directly. Every result links back to its film, director and cinematographer.

Can software query the database directly?

Yes. A REST API with 11 endpoints and no rate limits, an MCP server with four tools for AI agents, and an npm SDK (npm install framethrower-ai). A free account gets $2 of credits that work through the API from the first call — no card and no plan required. Endpoints return metadata, thumbnails, colour palettes and deep links, never raw image bytes.

Is it free to use?

Signing up is free and gets you $2 of credits with no card. After that it is usage-based: $1 = 1,000 credits, a search costs 2 credits, so $1 buys 500 searches. Credits never expire. The film, director, cinematographer, technique and colour-palette pages are public and need no account at all.

Who owns the images?

Not us. Frames are shown at reduced resolution as references for commentary, education and study, and rights remain with their respective owners. We license the measurements we compute — colour palettes, counts, rankings — under CC BY 4.0, but that licence covers the measurements only and never the frames themselves.