Former OpenAI Researcher Releases Flipbook Prototype: Direct Pixel Generation Using AI Video Models

According to monitoring by Beating, former OpenAI researcher Zain Shah and his team have released Flipbook, an experimental prototype that directly generates screen pixels using AI models, replacing traditional web technologies like HTML and CSS. Each ‘page’ viewed by users is an AI-generated image, and clicking on any area of the image generates a new image for further exploration. The entire interface is free of HTML code, fixed links, and predefined buttons, with even the text represented as pixels in the image. The video mode is based on the open-source DiT (Diffusion Transformer) video generation model LTX Studio from the Israeli company Lightricks, optimized to stream in 1080p at 24fps to users’ screens via WebSocket, with a backend powered by Modal Labs’ serverless GPU. Shah mentioned that the current functionality of Flipbook is limited, and the team is designing around visual explanations, but it showcases a larger direction: as models become more accurate and stateful, it could expand into structured UIs, including programming scenarios. Shah previously worked on AI and robotics research at OpenAI, later served as a creative technology expert at Samsung, and is also a YC S13 alumnus. Team members include former engineers from Humane and Slack, Eddie Jiao, and former Apple engineer Drew O’Carr.