FrameThrower Turns Film Stills Into Searchable References

A filmmaker may know the image they want before they know the film that contains it. It might be a night exterior with wet streets, a close-up lit from one side, or a muted colour palette at the end of a scene. Finding examples usually means remembering titles, scrubbing through films, or collecting disconnected screenshots in folders.

FrameThrower is built around that search problem. It presents a library of film stills that can be searched using descriptions of visual qualities, then saved into lookbooks for pitches, production and study. The useful idea is not simply having more movie screenshots in one place. It is being able to start with a visual intention rather than a film title.

Search by the image in your head

FrameThrower says its frames are indexed by attributes including lighting, lens character, shot size, colour grade, time of day and the people in the shot. That gives the service a different starting point from a conventional film database. A search such as “neon rain at night”, shown on the landing page, is closer to how a director or cinematographer might describe a reference during early planning.

The result, in theory, is a way to research cinematography references without first knowing which directors or films to look for. You can explore lighting and composition, study how a particular shot size is used, or collect images around a colour and mood. The product describes its library as covering millions of shots and thousands of titles, although those are the developer's claims rather than figures independently verified here.

That distinction matters. AI search can reduce the time spent finding a useful starting point, but it does not replace visual judgement. A description can be interpreted in several ways, and a technically relevant frame may still be wrong for a particular story, budget or audience. FrameThrower appears most useful as a discovery layer, not as an authority on what a scene should look like.

From scattered references to a working board

The second part of the product is organisation. Search results can be saved to lookbooks, where references are arranged by scene, sequence or overall look. The landing page also describes a workflow in which a script is used to create a scene-by-scene visual pitch from existing film frames.

That could help during the stage when a filmmaker needs to explain an idea before there is footage to show. A board can give a producer, director of photography, designer or client a shared visual language. The service says boards can be shared by link and exported as PDF files or ZIP files, which makes them suitable for pitch documents and production references rather than only private browsing.

This is a more specific use than making a moodboard for general inspiration. A useful board needs structure. Grouping images by scene can expose whether the proposed visual language is consistent, while grouping them by lighting or shot size can help a crew discuss decisions in concrete terms. FrameThrower gives the reference-gathering process a place to happen alongside that arrangement.

The browser is only one entry point

FrameThrower runs on the web and also offers an API. The developer describes a REST API, an npm SDK, an MCP server for Claude and ChatGPT, and a ComfyUI node. The page shows developers creating an API token, searching through the SDK or REST, and allowing an agent to call tools for search, crafting, similar frames and frame details.

Those options make the product relevant to teams that already have a creative workflow built around scripts, assistants or image tools. A developer could use the API to bring reference searches into another application, while a user of an AI coding or creative workflow could access the MCP server instead of opening the library manually. The submission also says the service has a free plan with a paid upgrade.

For someone who only needs a handful of references for one project, these integrations may be more infrastructure than they need. The main value remains the visual search and lookbook workflow. The API matters when references need to move through a larger process, not simply when a person wants to browse stills.

Important limits around the stills

FrameThrower is a reference tool, not a source of production-ready image assets. Its landing page says stills are shown at reduced resolution for commentary, education and reference, and that rights remain with their respective owners. That limits how the images can be used in a commercial project. A PDF lookbook can communicate an intended style, but it does not grant permission to reuse the frames in advertising, a final cut or other published work.

The platform list is also narrow. It is available on the web and through an API, with no desktop or mobile application listed. That is not necessarily a problem for browser-based research, but it may matter to crews that work offline or expect a dedicated production app. The quality of the search results, the completeness of the library and the usefulness of generated scene boards will also need to be judged in actual projects. The supplied material does not provide independent tests of those points.

FrameThrower is aimed at filmmakers, cinematographers and advertising creatives who spend time finding visual references and turning them into a coherent pitch or production board. That audience is a reasonable fit for the features described. It is less suitable for anyone looking for licensed footage, final artwork, offline software or a replacement for their own creative decisions. Its strongest promise is narrower and more practical: help people move from “I want this kind of image” to a usable set of references without starting from a known film title.