Image intelligence workflow
Understand and segment any image.
Paste a public Instagram post or Reel URL, choose an image from your device, or paste one from the clipboard. Post images and up to 10 detected Reel start frames are selected by default and run through understanding and segmentation in parallel.
Pipeline chat
Original and segments
Recover a missed segment
Mark one or more points inside the same missed object, retrieve candidate masks, then map each useful result to an entity.
Click the image to place the first point.
Recovered candidates
Image description
Entities
Select mask thumbnails to choose the product segments you want to validate, reconstruct, or search.
Select one or more mask thumbnails to extract crops.
Canonical organization
Canonical entities
Drag a segment onto another entity, or use Move to choose an A–Z destination. Delete removes a segment from canonical entities without deleting its source asset. Changes last for this page session.
Cross-image segment workflow
Segment workspace
Validate, reconstruct, and search extracted segments from every image in one place. Select up to 10 prepared segments for independent index searches.
Visual product search
Visual search results
Review grounded product matches, dimension and pricing sources, and the cost of each search.
Private Chinese catalog
Chinese furniture matches
Compare a pre-canonical crop against ABest Furniture products using interleaved product text and photography.
Selected vector indexes
Index matches
Each prepared segment is searched independently. Successful matches and retryable failures stay grouped by source segment.
Session moodboard
Saved finds
Add results to compare products here. The moodboard lasts for this browser session only.