SAM 2 introduces a single, Apache-2.0-licensed checkpoint capable of zero-shot object segmentation across stills and footage. Users can prompt with a click, box or rough mask; the model tracks selected items through occlusions, motion blur and scene cuts in real time. Optimised for CPU and GPU deployment, SAM 2 integrates into mobile apps, annotation pipelines and research workflows, reducing manual masking hours to seconds without additional fine-tuning.
Your personal, proactive Google AI assistant that simplifies work, school and everyday life right on your phone
✓Free
Fastest gateway to Gemini multimodal models with 2 M token context, caching and search grounding
✓FreeGenerate realistic, physics-aware videos up to one minute from text, images or existing footage
