Text prompts
Describe a visual concept with a short phrase and find matching objects across an image.
SAM 3 model guide
SAM 3 Model: What It Does and How to Try It
SAM 3 is Meta's promptable model for detecting, segmenting, and tracking visual concepts in images and videos. It accepts text, image examples, points, and boxes, then returns masks for matching objects.
Official SAM 3 resources
Start with the primary sources for research details, code, checkpoints, and supported workflows.
How SAM 3 can segment images
Points and boxes
Select a specific object interactively and refine its mask with positive or negative points.
Image examples
Provide an example object to identify other instances that match the same visual concept.
Video tracking
Track selected concepts and object instances across video frames.
Choose how you want to try SAM 3
Use the private in-browser playground for point and box selection, or open the cloud workspace for text prompts, saved history, and production workflows.
SegmentLens is an independent product and is not affiliated with or endorsed by Meta. SAM 3 and related marks belong to Meta Platforms, Inc.