Reelary
Free · runs in your browser · no upload

Find a moment in a video by describing it

Search inside a video by describing what you are looking for. Drop in an MP4 or MOV, type "someone opens a laptop" or "sunset over the water", and jump straight to the moments that match. Free, private, and the video never leaves your computer.

Try:

Your video is read by this browser tab and never sent anywhere, and neither is what you type. The first run downloads the AI model from Hugging Face and caches it, so later runs start faster and work offline.

How searching a video by description works

  1. Read the frames. The tool takes a frame about every second and turns each one into a list of numbers that captures what is in the picture, using CLIP, an image-text model from OpenAI.
  2. Read your words. Your search is turned into numbers the same way, in the same space, so "a red car" lands near frames that show a red car.
  3. Compare. Every frame is scored against your words at once. Neighbouring frames that score almost as well as the best one are joined into a single moment with a start and end time.
  4. Jump. The best moment plays straight away; click any other result to jump to it.

Why not just scrub through the video?

For a two-minute clip you can. For a 20-minute interview, a screen recording or an hour of drone footage cut into parts, scrubbing means watching most of it again. File names and transcripts do not help when the thing you remember is visual: the shot with the red umbrella, the moment the sun came out, the bit where the whiteboard is in frame. Describing it is faster than watching for it.

Writing a good search

  • Describe the picture, not the story. "Two people shaking hands in an office" works better than "the deal gets signed".
  • Add the setting or the light when a subject appears many times: "dog on the sofa" versus "dog on the beach".
  • Try a second wording if the first results are close but not right. Searching again is instant.

When the desktop app is the better tool

This tool searches one video. The moment you are looking for is usually in one of hundreds: last year's trip, a folder of client rushes, a drive of b-roll. The Reelary desktop app reads every video and photo you point it at once, in the background, and then the same kind of search covers all of them, along with what was said and the text on screen.

Results are clips with exact start and end times, so you can drop them onto a timeline and export a cut, or hand the timeline to Premiere, Final Cut or DaVinci Resolve. See how local AI video search works.

The desktop app does this across every video you own at once, then lets you cut the results into a video. Reelary is free to download for Mac and Windows; indexing and search have no file limit.

Download Reelary free How local search works

Questions

How do I find a specific moment in a long video?

Drop the video into the tool above and type what happens in the shot you want, in plain words: "a dog jumps into the pool", "close-up of hands on a keyboard", "the crowd cheering". The tool shows the best-matching moments with timestamps and plays the first one. Change the words and search again; it does not need to re-read the video.

Can AI search inside a video for a specific scene?

Yes, by looking at the picture. This tool looks at a frame about every second across the whole video and compares each one with your words, so it can point at the exact stretch you mean instead of summarising the video as a whole.

Does it search what people say?

No, it searches the picture. To find a line someone said, transcribe the video with the free video to text tool and search the transcript. The Reelary desktop app searches both the picture and the speech of every video you own.

Is my video uploaded?

No. Your browser decodes the video and the CLIP model runs in this tab, on your graphics card through WebGPU or on the CPU. The model is downloaded once from Hugging Face and cached. What you type is never sent anywhere either.

What kind of searches work best?

Describe what is visible: objects, people and what they are doing, places, light, colours, camera angles ("aerial shot of a coastline", "person walking in the rain at night"). It does not know who a specific person is, cannot read small text and sees single frames, so it is weaker at actions defined by movement, such as "the moment the ball goes in".

What are the limits?

Videos up to 20 minutes. The tool reads one frame per second for videos up to 10 minutes, and 600 evenly spaced frames for longer ones. The model is a one-time download of about 150 MB on the CPU or about 240 MB with WebGPU. On a recent laptop each frame takes about a tenth of a second with WebGPU and a quarter of a second on the CPU, so a 10-minute video is read in one to three minutes; every search after that is instant.

Which model does it use?

OpenAI CLIP ViT-B/32, released under the MIT license, run with Transformers.js. The Reelary desktop app uses larger image-text models of the same kind across your whole library.

More free tools

AI video describer Splits a video into scenes and writes a plain-English description of each one with a vision model running in the browser. Exports TXT, SRT or WebVTT. No upload. Video background remover Cuts the person out of a video with an AI matting model in the browser and puts them on transparency, a solid colour or a blurred background. Keeps the sound, no watermark, no upload. Remove logo from video Mark the logo, date stamp or caption in your own footage and an inpainting model fills that area on every frame, keeping the sound. Best for marks that stay in one place. No upload. Video to text Transcribes speech in a video or audio file with OpenAI Whisper running locally in the browser, with timestamps. Exports TXT, SRT and WebVTT. No upload, no sign-up. Caption generator Generates captions for a video with Whisper in the browser, splits them into short readable lines, lets you edit them on a live preview and exports SRT, VTT or a video with the captions burned in. No upload. Extract frames from video Finds the scene changes in a video, or samples it every N seconds, and saves full-resolution JPG or PNG stills as a ZIP. Runs in the browser, no upload. Add caption to GIF Puts a caption on an animated GIF or a short clip: white caption bar, top and bottom meme text, or subtitles. Live preview and GIF export in the browser, no watermark.

Published October 19, 2026 · Models we use