Search screen text and speech
Find visible text or spoken words, then jump to the match.
Search the two parts of the recording
Screen text comes from recognizing text visible in captured frames. Transcript text comes from recognizing the recorded audio. Together, they help recover a moment even when you remember the wording but not the time.
- Open the recording in the viewer.
- Enter a useful phrase in the search field.
- Use the available search options and source filters to refine the results.
- Select a result or timestamp to jump to that moment.
- Play a little context to confirm it is the moment you need.
Choose a search mode
The viewer provides Fuzzy, Exact and Semantic modes. Use Exact for specific wording, Fuzzy when recognition or spelling may differ, and Semantic to look for related meaning when its local embedding index is available. Semantic results should still be checked against the recording.
Try the Northstar example
In the demo, searching weekly project digest connects a visible preference with the spoken explanation. The result leads back to the recorded setting rather than a separate written note.
If a result is missing
Check enrichment status first. The words may not have been legible on screen or clear in the audio. Try a distinctive shorter phrase, review the transcript, or use the library’s search-audit tools to inspect coverage.