Memories.ai

Memories.ai is an AI video intelligence platform built around what it calls a Large Visual Memory Model, designed to give AI systems persistent, searchable memory of video content rather than analyzing clips in isolation. The platform turns raw footage into structured visual data covering frames, faces, scenes, and audio, then organizes that data into a searchable knowledge layer so users can search, summarize, and act on large volumes of video quickly. It targets a range of industries: media and entertainment companies use it for searchable archives and real-time clip generation, security and safety teams use it for real-time threat detection and cross-camera person tracking, and robotics developers use it to give robots and wearables visual memory. The company was founded by former Meta Reality Labs researchers and describes itself as built by former scientists from Meta, Google, and Nvidia. Memories.ai offers a Video Intelligence API and a Visual Search API for developers, supports edge, on-premises, and cloud deployment, and integrates with hardware and platform partners including Microsoft, NVIDIA, Samsung, Lenovo, Qualcomm, and Ring, reportedly powering over 100 million cameras. Pricing is usage-based, billed per inference and per storage. As a young company built around a novel model architecture, its claims about surpassing existing approaches are not independently verifiable from the site alone.