Analysis updated 2026-05-18
Build a private lookbook of cinematic stills for a film or commercial project.
Search a personal archive of movie frames by visual look, mood, or technique.
Collect and organize reference footage into moodboards with notes and stickies.
| gianluca-improta/cinekive | huey1in/windsurfx | keon/jepa | |
|---|---|---|---|
| Stars | 97 | 97 | 97 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 4/5 |
| Audience | general | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Docker Desktop is optional on Windows and Mac but still required on Linux, first search may download an ~800MB model.
Cinekive is a local, searchable visual archive for filmmakers who want to collect and organize cinematic stills without relying on someone else's online library. You can drop in a film, a folder of stills, or a video URL, and the app finds the standout frames, tags them by craft, and lets you find any frame later by look, director, technique, or mood, instead of scrubbing through a timeline. It is inspired by existing sites like FilmGrab, EyeCandy, Flim, and Kive, but built to run entirely on your own machine so your archive stays private. The app is meant for editors and directors building lookbooks, commercial and social media teams who need client safe references, and cinephiles or students collecting stills they have the right to use. It ships as a desktop app for Windows, Mac, and Linux, or it can run in a browser. Setup is meant to be simple: download the release for your operating system, pick a library folder, and start. On Windows and Mac, Docker is optional, since the app can download a native engine pack the first time it runs, though installing Docker Desktop makes setup faster. On Linux, Docker Desktop is currently still required. Under the hood, Cinekive uses a local AI model called SigLIP for visual search, along with yt-dlp for pulling in video from URLs, and it can optionally use Ollama to generate craft tags. Archive sources include mirrors of FilmGrab, EyeCandy, ShotDeck, MovieStillsDB, and StillsLab. Beyond search, the app includes moodboards, which are infinite canvases where you can drag in project clips, notes, stickies, and audio or media links, and organize them into named concepts and stacks. The interface supports several languages, including English, Chinese, Spanish, French, German, and Japanese. Everything stays local by default, with no cloud account needed, and your media folder is never stored inside the project's code. There is also a local HTTP API meant for automation or multi agent workflows that want to interact with the archive programmatically.
A local, private, searchable archive app for filmmakers to collect, tag, and organize cinematic stills and video frames by look, mood, and technique.
Mainly Python. The stack also includes Python, FastAPI, Next.js.
Released under the MIT license, so it can be used, modified, and redistributed freely, including for commercial purposes.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.