Search Examples
An interview quote, a product shot, or a person from a previous shoot can become a search query. These examples show how AI video search uses the indexed content to find matching moments.
Speech becomes a transcript with timestamps. Speaker diarization separates the speakers.
"quarterly goals"Find a passage containing that phrase.
Transcription search guideDetected objects are linked to moments in the footage.
laptopReview shots containing a detected laptop.
Object detection search guideScene descriptions make the setting and action searchable.
coastline at sunsetLook for matching scenes across shoots.
Scene descriptions search guideFace clustering groups appearances of the same person. Face and voice matches can be assigned a name.
@SarahFind footage linked to Sarah after naming her.
People search guideHow It Works
Start with footage from a current project. You can index more clips later and search them alongside the existing library. Processing time varies with your files, device, and connection.
Add a folder from your drives and scan its files. FrameQuery catalogs paths, durations, formats, and camera metadata. Select the footage you want to index.
FrameQuery extracts sampled frames and audio on your device. Those are sent for cloud analysis. Your original video files stay on your drives.
Cloud analysis produces transcripts with speaker diarization, object labels, and scene descriptions. Face and voice recognition run on your device. Results are associated with the relevant moments in each clip.
FrameQuery stores the analysis results and search index on your machine, using Tantivy for keyword matching. The extracted frames and audio sent for analysis are discarded after processing.
Search a phrase, describe a shot, or filter by a named person. Open a result at its matching timecode. Indexed footage remains searchable offline, with no per-query charge.
Privacy
The indexing servers see only extracted frames and audio, and return structured metadata. Biometric data is generated and stored exclusively on your device.
| Data | Location | Cloud processing |
|---|---|---|
| Original video files | Your drives | No |
| Extracted frames and audio | Your machine (transient) | Sent for analysis, discarded immediately after |
| Transcripts and metadata | Local search index | Analysis results returned to your device |
| Face and voice embeddings | Local encrypted storage | Never |
| Search index | Your boot drive | No |
Choosing a Tool
Check whether you need search within an editing project, a library on your drives, or a shared media asset manager. Compare supported files, result timecodes, and access for the people using your footage. Our guide to the best video search software covers those options. You can also compare FrameQuery with Iconik and Frame.io. For professional camera files, check our format compatibility list.
For an application that needs to submit footage for analysis and consume the resulting metadata, see the video search API.
FAQ
Video indexing software analyzes video content and records searchable information such as transcripts, scene descriptions, objects, and people. FrameQuery connects those results to timecodes in your footage. You can then search across the indexed library.
Processing time depends on the duration and format of your footage, the speed of your device, your connection, and the analysis being performed. Start with a representative batch from a current project before queuing a large archive. Completed clips become searchable as their indexing finishes.
FrameQuery sends extracted frames and audio for cloud analysis. Your original video files stay on your drives. The extracted data is discarded after processing, and the search index is stored locally. Initial cloud analysis requires an internet connection.
The size depends on the amount of footage, transcript length, scene metadata, and stored search embeddings. Thumbnails and playback caches also use disk space. Index a representative batch and check its storage use before estimating a full archive. The search index is separate from the original video files.
Scanning catalogs files and their metadata, such as duration, codec, and resolution. AI indexing adds information about the footage itself: dialogue, scenes, objects, and people. Catalog a source folder first, then choose the footage that needs content analysis.
FrameQuery supports 50+ codecs and formats, including R3D, BRAW, ProRes, ProRes RAW, CinemaDNG, XAVC, MXF, DNxHR, H.264, H.265, and AV1. Format support depends on the codec and operating system. Check the compatibility page for the files you use.
Searching previously indexed footage is free on every plan, including offline. Paid indexing starts at $19/month for 10 hours of footage. You can keep searching your existing index after a paid plan ends. Keep the source files available to play or export the footage.