Shofo

Companies building the next generation of multimodal AI models need large volumes of diverse, high-quality training data such as computer use logs, talking head footage, speech, gameplay, and expert physical labor video. This kind of data, especially with full input logs or detailed transcripts, is not easily available and must be sourced or produced at scale. Shofo's page is addressed to AI labs that need this data to power model development.

Shofo supplies millions of hours of proprietary data across several categories: expert computer use data (CAD, Photoshop, researcher workflows) with full screen, keyboard, and mouse logs; talking head footage with transcripts filterable by attributes like objects or hands covering faces; speech and audio data including podcasts and interviews with transcripts in single or dual speaker formats; video game data with full gameplay and input logs; and expert physical labor footage such as electrician, surgical, and carpentry work. The data is offered through a self-serve platform where prospective customers can browse datasets, and the company also allows direct contact to request a sample.