Biography

Bill Peebles is a computer vision researcher at OpenAI who serves as one of the research leads for Sora, the company’s text-to-video diffusion transformer. The Sora launch credits Peebles and Tim Brooks with steering the research program and coordinating the wider engineering group responsible for the model’s production deployment.

Leadership of the Sora program

  • Text-to-video generation (2024–present). Sora’s public reveal highlighted that the model can generate up to one minute of high-fidelity video from natural language prompts, extend existing clips, and animate still imagery—capabilities Peebles oversees as part of the program’s technical roadmap.
  • Diffusion transformer architecture. Under Peebles’ leadership, Sora adopts a transformer-based diffusion approach that reasons over video “patches,” enabling consistent characters and camera motion across long sequences.
  • Safety and staged access. The Sora rollout includes red-teaming, policy enforcement, and provenance markers before wide release as part of the program’s staged deployment process.

Generative modeling research

Peebles’ prior academic work focuses on scaling diffusion transformers for image synthesis and generative modeling of neural network checkpoints. His papers “Scalable Diffusion Models with Transformers” and “Learning to Learn with Generative Models of Neural Network Checkpoints” explore transformer backbones for diffusion processes and conditioning updates on desired objectives.

Collaborative team structure

Peebles co-leads Sora with Tim Brooks while working closely with systems lead Connor Holmes and a cohort of OpenAI researchers specializing in video generation, rendering, and infrastructure. This structure pairs research leadership with domain experts in safety, communications, and legal operations to support Sora’s gradual deployment and policy alignment.

Focus areas

  • Scaling multimodal diffusion transformers for long-form, instruction-following video
  • Maintaining character and environment coherence across complex prompts
  • Integrating safety review, classifier tooling, and provenance metadata into video generation workflows

These areas reflect Peebles’ role in OpenAI’s video generation research program.