Former OpenAI researcher releases Flipbook prototype: skip HTML, use AI video model to directly generate every pixel

robot
Abstract generation in progress
ME News, April 23 (UTC+8), according to Beating monitoring, former OpenAI researcher Zain Shah and his team released Flipbook, an experimental prototype that uses an AI model to directly generate screen pixels, replacing traditional web technologies such as HTML and CSS.
Every "page" the user sees is an AI-generated image. Clicking on any area of the image generates a new image to continue exploring. The entire interface has no HTML code, no fixed links, no predefined buttons, and even the text is pixels in the image.
The video mode is based on the open-source DiT (Diffusion Transformer) video generation model LTX Studio by Israeli company Lightricks. After optimization, it can stream in real-time at 1080p 24fps via WebSocket to the user's screen, with a backend connected to Modal Labs' serverless GPU.
Shah stated that Flipbook currently has limited functionality, and the team designed it around visual explanation, but it demonstrates a larger direction: as models become more accurate and stateful, it can be extended to structured UI in the future, including programming scenarios.
Shah previously worked on AI and robotics research at OpenAI, then served as a Creative Technology Expert at Samsung, and is also a YC S13 alumnus. Team members also include former Humane and Slack engineer Eddie Jiao, and former Apple engineer Drew O'Carr.
(Source: BlockBeats)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments