Meta's AI Sees All: New SAM 3 Powers Smarter Editing

Meta has unveiled the third generation of its Segment Anything Models (SAM 3), a series of AI systems designed for sophisticated visual intelligence. These models specialize in object detection, capable of precisely identifying and segmenting elements within images and videos. SAM 3 can distinguish specific items, such as all elephants in a photo, or respond to text prompts like "highlight red caps."

The SAM 3 models are not generative AI but are intended to significantly enhance content editing capabilities across Meta's suite of platforms. Users can anticipate more refined editing tools in applications like Instagram Edits and the AI video app Vibes. Additionally, the Facebook Marketplace "view in a room" feature will leverage SAM 3 to provide realistic previews of items like furniture in a user's home.

Beyond consumer applications, SAM 3 has demonstrated its utility in scientific research. Meta partnered with conservation organizations, utilizing the AI to analyze vast amounts of wildlife video footage, thereby improving animal identification for monitoring and conservation efforts. This advancement highlights the growing integration of AI in understanding and interacting with the physical world.