FrameMap
FrameMap Beta
FrameMap is a visual exploration tool that projects every indexed frame in your library onto a 2D canvas, organized by visual similarity. Frames that look alike appear close together; unrelated frames are far apart.
What it shows
Each dot on the map represents a single frame from one of your videos. The position of each dot reflects how visually similar it is to every other frame — computed from the same embeddings used for search.
Frames are automatically grouped into clusters. Each cluster represents a visual theme across your library: all your outdoor scenes, all your whiteboard moments, all your interview talking-head shots.
How to use it
FrameMap is available in the Pureframe app at console.pureframe.ai/framemap.
- Hover over a dot to see a frame preview and video name
- Click a dot to pin it to the preview strip at the bottom
- Double-click to open the video at that exact timestamp
- Lasso select a region to inspect multiple frames at once
- Use the minimap in the corner to navigate large libraries
- Filter by collection using the left panel
Clusters
HDBSCAN clustering automatically groups frames into named clusters. The right panel shows all clusters with frame counts. Click a cluster to highlight all its frames on the map.
Frames that don’t belong to any cluster are shown as “Uncategorized” — these are typically unique or transitional frames.
Recomputing
FrameMap is computed and cached when you first visit it. Click Recompute in the left panel to regenerate it after uploading new videos.
FrameMap is in beta. Cluster quality improves as you add more videos to your library — at least 50 indexed videos is recommended for meaningful groupings.
API access
FrameMap data is available via GET /v1/framemap. The response includes every frame’s 2D coordinates, cluster assignment, and thumbnail URL.
Filter by collection:
