How to search inside video files
Your computer can find a video by its name, but not by the sentence someone said at minute 14. To search inside a video you first have to turn its speech, on-screen text and pictures into something searchable.
The short answer
Video files are opaque to ordinary search. To search inside them you need an index built from the content: a transcript of the speech (with timestamps), text read off the frames, and ideally a description of what each scene looks like. Once that exists, a search box can take you to the exact moment.
You can build a rough version by hand with free tools, or use an app that does all three on your own computer and links every result to its timestamp.
Why Spotlight and Windows Search can't do it
Spotlight on macOS and Windows Search index file names, dates, tags and the text inside documents like PDFs. They do not transcribe the audio track of a video or read the words on a slide that appears in it. A recording called Screen Recording 2025-03-02.mov is, as far as they're concerned, just a file name and a length.
That's why renaming files and keeping tidy folders only gets you so far. You'd have to remember which file, and then still scrub through it to find the part.
Three ways to make video searchable
Each approach catches a different kind of moment, so it helps to know which one matches how you remember things:
- Search the speech. Transcribe the audio, keep the timestamps, and search the text. This is the biggest win for interviews, lectures, meetings and podcasts. Keyword search finds exact words; semantic search also finds a passage when you only remember the idea.
- Search the on-screen text. Sample frames and run OCR so slide titles, captions, signs and error messages become text. Essential for screen recordings and presentations.
- Search the picture. Visual search matches a description like "a dog running on a beach" against the frames themselves. This is how editors find footage that has no dialogue at all.
The do-it-yourself route
If you only have a few files, a manual workflow works. Run each video through a transcription tool that outputs subtitles (SRT or VTT files carry timestamps), then search those text files with your normal file search or a text editor. When you find the line, the timestamp next to it tells you where to jump in the player.
The limits show up quickly: you're maintaining a pile of sidecar files, you only catch exact words, and anything that was shown rather than said stays invisible. Uploading to a cloud transcription service removes some of the work, but it means sending every recording to someone else's servers and usually paying per minute.
How to do it in MediaFind
MediaFind builds all three indexes locally. Nothing is uploaded; the models download once and then work offline.
- Install MediaFind (the Mac app is the main build; Windows and Linux are in beta) and pick a transcription model sized to your machine on first run.
- Add the folder that holds your videos. MediaFind transcribes each file with word-level timestamps, reads on-screen text, extracts keyframes and labels speakers, all on your computer.
- Type what you remember in plain words, for example the part where she explains the pricing change. Results are ranked moments with exact timestamps.
- Narrow by how you remember it: on-screen text, visual scene, speaker, and other search modes are all in the same search box.
- Click a result to play from that second, or cut it out as a clip. If you want an answer rather than a clip, use Ask, which replies with the sources and timestamps it used.
Tips and limitations
Indexing takes time on the first pass, since every file has to be transcribed and analyzed; files become searchable one by one as they finish. While the app is open it re-checks your added folders about once a minute and indexes new or changed files after they finish copying, so you don't have to re-add anything.
Transcripts are only as good as the audio. Heavy crosstalk or music under speech will lower quality, and adding a names and terms list helps with unusual names. The free plan indexes up to 10 media files, which is enough to test on your own recordings before deciding whether you need more. For a deeper walkthrough, see finding a spoken word in video and searching a whole video library.
Frequently asked questions
Can I search inside a video without uploading it anywhere?
Yes. You need a tool that transcribes and indexes on your own computer. MediaFind does this on-device: after a one-time model download, indexing and search work with no internet connection.
Does Spotlight search the words spoken in a video?
No. Spotlight indexes file names, metadata and the text in documents, but it does not transcribe the audio in video or audio files.
What if I don't remember the exact words?
Use semantic search, which matches meaning rather than exact wording. Describing the idea, such as the part about the budget, can find a passage even if nobody said those words.
Which video formats can MediaFind search?
MP4, MOV, M4V, MKV and WebM for video, plus MP3, M4A, WAV, AIFF, FLAC and OGG audio and common photo formats.
Search your whole library on your own computer.
Free for up to 10 files, with a 7-day Pro trial. No account, nothing uploaded.
Download for macOS View pricingKeep reading
How to find b-roll by describing the shotType "aerial of a coastline at sunset" and get the clips, even though nobody logged them and nobody speaks in them. What is semantic search?
Searching by meaning instead of matching words, and why it helps most when you can't remember exactly what was said.