Memories.ai – Video Data Infrastructure, Action Indexing & Dense Annotation
#Quasa #QUA #Memoriesai
Memories.ai develops video data infrastructure for robotics and other physical AI applications. The company was founded by Shawn Shen and Ben Zhou, who previously worked at Meta. Its Video Datalake collects first-person footage of real tasks, indexes actions and supports annotations using a customer's own labels. Outputs include hand-specific captions, action segments and timestamps delivered as JSON. Teams can source footage from the company's collection or submit their own video for indexing and annotation. Developer access includes a REST API and MCP server. API pricing is usage-based, with enterprise deployment options; current pricing is listed on the official website. Dataset and custom capture terms should be confirmed separately.
𝗖𝗢𝗥𝗘 𝗦𝗧𝗥𝗘𝗡𝗚𝗧𝗛𝗦
• First-person video for robotics training
• Actions indexed for each hand
• Annotation using customer-defined labels
• Searchable video through developer tools
• Existing footage and custom capture
𝗜𝗗𝗘𝗔𝗟 𝗙𝗢𝗥
Robotics labs and engineering teams needing searchable, annotated video of human tasks.
𝗛𝗜𝗚𝗛𝗟𝗜𝗚𝗛𝗧𝗦
• JSON annotations with segment timestamps
• REST API for video retrieval
• MCP server for agent access
• Usage-based API billing without seat fees
• Enterprise on-premises deployment option
𝗣𝗢𝗧𝗘𝗡𝗧𝗜𝗔𝗟 𝗖𝗢𝗡𝗦𝗜𝗗𝗘𝗥𝗔𝗧𝗜𝗢𝗡𝗦
• Confirm dataset licensing and consent terms
• Validate annotation quality on representative tasks
• Check coverage of required environments
• Confirm API version and applicable rates
• Request custom capture costs and timelines
𝗢𝗩𝗘𝗥𝗔𝗟𝗟 𝗩𝗘𝗥𝗗𝗜𝗖𝗧
3.8/5 stars. Consider it for sourcing and structuring video for robotics research. Confirm sample quality, data rights and total costs first. Earn a QUA reward by reviewing on Quasa.io too!
𝗚𝗘𝗧 𝗦𝗧𝗔𝗥𝗧𝗘𝗗: https://quasa.io/projects/memoriesai























