Skip to content
← All guides

Find Someone in a Long Video Without Watching It — Get Exact Timestamps

Scrubbing a long video by hand is not a search. Sample frames locally, match them against one reference photo, and get merged time ranges that tell you exactly where to put the playhead.

Published 2026-10-08 5 min read

You have a video file and you need to find the moment a specific person appears in it. The obvious approach — playing the video and watching for their face — works for a five-minute clip. For anything longer, it becomes the most tedious task in your day.

The practical alternative is to search the video by face: give a tool one photo of the person, let it scan the footage, and get back the exact moments where they show up. This guide walks through how that workflow works and how to get the best results.

Why finding someone in video is hard

Video is not searchable the way text is. There is nothing to grep. The faces are there, but they are buried inside a stream of frames — 24 to 30 per second, each one a full image.

The only way to find someone without watching the whole thing is to automate the looking: sample frames at a regular interval, check each one for the face you are looking for, and turn the hits into something you can act on — timestamps.

That is exactly what SnapByFace does.

The workflow

1. Get a reference photo

You need one clear, front-facing photo of the person. It does not need to be high resolution — it needs to show their face clearly. A passport-style photo works well. A selfie in good light works too.

Tips for a better reference photo:

  • Front-facing — faces turned at an angle encode differently and may not match as well.
  • Good lighting — harsh shadows or heavy backlight reduce the quality of the face encoding.
  • Recent — if the person’s appearance has changed significantly, use a more recent photo. You can add multiple reference photos for the same person if needed.

2. Add your video files

Point SnapByFace at the folder (or folders) containing your video. It reads MP4, MOV, AVI, MKV, M4V, WebM and TS. It walks directories recursively, so a deeply nested folder tree is fine — just point it at the root.

You can add photos and video to the same index. If the person also appears in photos in those folders, one search finds them across both media types.

3. Start the index

The indexing process samples frames, detects faces and builds the search index. It runs in the background and you can:

  • Pause it if you need your machine’s resources for something else.
  • Resume from where it stopped — progress persists.
  • Cancel and restart if you pointed it at the wrong folder.

After an unexpected quit (crash, power loss), it resumes from where it left off. You do not lose progress.

4. Search and jump to the moment

Once indexed, searching is instant. You select the person from your people library and SnapByFace returns every moment they appear in the footage, grouped into time ranges.

Each result shows a time range — “02:14–02:34” — and clicking it jumps the preview straight to that second. No more dragging the timeline and squinting at faces.

How long does it take?

The work scales with how long the video is, not how big the file is. An hour of footage means about 3,600 frames to examine at the default one-per-second interval.

Two things keep this practical on large collections:

  • Incremental re-indexing. Adding new footage does not re-analyse what is already indexed. Only new or changed files are processed.
  • Fast search at any size. The search itself is near-instant even across hundreds of thousands of faces. The bottleneck is the initial indexing, and that only happens once.

If you want to see how it behaves before committing your whole archive, index one folder first and check the top hits against your own expectations. That is the cheapest way to learn whether the default settings are right for your footage.

Tips for better results

Use the right similarity threshold

The default threshold (0.45) works for most cases. If you are getting too many false positives (people who look similar but are not the right person), raise it to 0.55 or 0.60. If you are missing the person entirely, lower it to 0.35.

Add a second reference photo

If the person looks significantly different in the footage compared to your reference photo — different hairstyle, glasses vs. no glasses, aged several years — add a second reference photo that matches their appearance in the footage more closely.

Adjust the sampling interval

The default (one frame per second) catches most appearances. If you are looking for a split-second flash — someone walking past a window quickly — tighten it to 0.1–0.5 seconds. For very long recordings where you just need to know whether someone shows up at all, a looser interval (5–30 seconds) finishes faster.

Common scenarios

Finding yourself in family video

You recorded a school play or a family gathering. You were behind the camera. Now you want to find the moments where you appear in the audience. One selfie as reference, point it at the video folder, and the search gives you every moment.

Locating a subject in interview footage

You have hours of raw interview footage and need to find every appearance of a specific interviewee across multiple files. Add all the folders, add one reference photo, and the search spans everything.

Finding a person in surveillance or security footage

You have security camera recordings and need to find when a specific person was on camera. The same workflow applies — though be aware that low-resolution cameras and extreme angles make the face encoding less reliable.

What stays on your machine

Recognition and search run locally. No video, no frame, and nothing about the faces in them is uploaded. Everything the app learns stays in ~/.snapbyface/. SnapByFace reads your media and never writes to it — no renaming, no moving, no re-encoding.

The only outbound data is an optional device statistics ping (machine code, app version, OS, language, activation state — no filenames, no footage, no search activity). It is on by default and can be turned off in Settings.

When the results will disappoint

  • Fast cuts and heavy motion blur produce frames where the face is not clear enough to match reliably.
  • Very small faces in frame (a wide shot of a crowd) are the hard case for any face search.
  • Heavy appearance change, or a reference photo that looks nothing like the person in the footage, lowers the score below the threshold.

Try it on your footage

The trial runs the full local search and shows up to 5 matches per person and 20 per search — enough to tell you whether it finds what you are looking for in your own material. During the trial you can start up to 100 analyses in total.

Try it on your own library

The free trial runs a full local search — no account, no upload.