Memories.ai emerged from stealth in July 2025, founded by former Meta Reality Labs researchers Dr. Shawn Shen and Enmin Zhou to tackle a fundamental limitation in today's AI: the inability to process and remember more than a few hours of video. The company describes its core technology as a Large Visual Memory Model (LVMM) — a system designed to give machines persistent, searchable visual memory that scales to enormous volumes of footage, reportedly toward as much as 10 million hours.
The platform positions itself as 'the fundamental visual memory layer for AI,' enabling capabilities like conversational video chat, clip search, and transcription over unlimited video context. In practice that means users can ask questions about hours of footage, locate specific moments instantly, and extract structured insight without manual review. The company claims its long-video understanding outperforms general-purpose models like Gemini and ChatGPT on extended context, and it emphasizes on-device processing and privacy compliance for sensitive deployments.
Memories.ai raised an $8 million seed round in July 2025 led by Susa Ventures, with participation from Samsung Next, Fusion Fund, Crane Venture Partners, Seedcamp, and Creator Ventures. Demand outpaced the team's initial $4 million target, reflecting investor appetite for foundational video-understanding technology. The company highlights partnerships and interest from major hardware players including Qualcomm, Samsung, NVIDIA, and Lenovo.
The target applications are broad and enterprise-oriented: security monitoring with threat detection and person tracking, sports analytics, media and production workflows, video marketing, and robotics, where persistent visual memory is essential for perception and decision-making. This breadth reflects the company's ambition to be infrastructure — a memory layer many products build on — rather than a single vertical app.
For businesses, Memories.ai addresses a problem that grows with every camera and recording: the world generates far more video than anyone can watch. By giving AI the ability to remember and reason over long footage at scale, the company aims to make vast video archives instantly queryable and actionable — a capability with implications across security, media, sports, and embodied AI.