Natural-language semantic search across project media directly in AI assistant

In large projects, finding a specific shot means relying on manual tagging, markers, or memory. This slows down rough cuts and makes projects hard to hand over to other editors.

Proposed solution: Add a search field in the Project panel that accepts plain descriptions (for example, "wide shot, hallway, person walking away" or "product close-up on white background"). Results would be ranked by relevance and show thumbnails with the matching timecode. Indexing should run locally in the background, with an opt-out for sensitive footage.

Use case: For a B2B product shoot with hundreds of clips, an editor could locate the right detail shot in seconds instead of scrubbing through bins, and a new team member could navigate the project without a briefing.