Tim Brooks

Tim Brooks is an AI researcher who was credited, alongside Bill Peebles, as a research lead on Sora, OpenAI’s text-to-video model announced in February 2024. OpenAI’s announcement page lists the two as research leads and Connor Holmes as systems lead.

Career

By his own account, Brooks worked on the camera software of Google’s Pixel phones and on video generation at NVIDIA, completed a PhD at Berkeley AI Research advised by Alexei Efros, then co-created Sora at OpenAI. He subsequently worked on world models and multimodal generation for Gemini at Google DeepMind, and states that he is now an AI researcher at Meta Superintelligence Labs working on multimodal generative models.

Research

Brooks’ doctoral work produced InstructPix2Pix, a method for editing an image from a written instruction. The paper — written with Aleksander Holynski and Alexei Efros — combined a language model and a text-to-image model to generate a training set of image-editing examples, then trained a conditional diffusion model on it that generalizes to real images and user-written instructions without per-example fine-tuning.

Relevance to political media

Brooks’ work sits in the line of research that made instruction-following image and video generation practical. Sora, the OpenAI system he is credited with co-leading, generates video from written prompts, and the difficulty of distinguishing such output from recorded footage is what gives the technology its bearing on political media.