Bill Peebles

As an undergraduate at the Massachusetts Institute of Technology (MIT), Peebles did research advised by Antonio Torralba. By his own account, he began researching video generation as an undergraduate and continued that work at Berkeley. In 2019 he co-authored “Semantic Photo Manipulation with a Generative Image Prior,” published at SIGGRAPH 2019, and “Seeing What a GAN Cannot Generate,” an oral paper at ICCV 2019.⁠[1][2]

Peebles then pursued a PhD at the Berkeley Artificial Intelligence Research lab (BAIR) at the University of California, Berkeley, advised by Alexei Efros, supported by the National Science Foundation Graduate Research Fellowship Program, and interned at FAIR, AdobeAdobeAdobe (Adobe Inc.) is a software company headquartered in San Jose, California, founded in 1982 by John Warnock and Charles Geschke. Known for PostScript, PDF and creative software such as Acrobat, Photoshop, Illustrator, Premiere Pro and After Effects, it moved its main creative products to Creative Cloud subscriptions from 2013 and expanded into digital marketing and generative AI; it is listed on Nasdaq under the ticker ADBE.Open the full entry Research and NVIDIANVIDIANVIDIA is an American semiconductor and computing technology company that provides graphics processing units, artificial intelligence infrastructure, networking interconnects, and computing software, and is listed on Nasdaq under the ticker NVDA.Open the full entry. In 2020 he was first author of “The Hessian Penalty: A Weak Prior for Unsupervised Disentanglement,” a spotlight paper at ECCV 2020.⁠[1]

In 2022 Peebles was first author of “GAN-Supervised Dense Visual Alignment,” an oral paper at CVPR 2022 and a best paper finalist.⁠[1]

In June 2022 Saining Xie began the research project that produced the diffusion transformer (DiT), with Peebles as his mentee while interning at MetaMetaMeta (Meta Platforms, Inc.) is a publicly listed American technology company, renamed from Facebook in 2021, that operates social communication, digital advertising, virtual reality, smart glasses and artificial intelligence businesses.Open the full entry’s AI research lab. On December 19, 2022, the two posted “Scalable Diffusion Models with Transformers” on arXiv, replacing the U-Net backbone commonly used in diffusion models with a transformer operating on latent patches; their largest model achieved an FID of 2.27 on the class-conditional 256×256 ImageNet benchmark. The paper was later accepted as an oral at ICCV 2023.⁠[3][4][1]

On May 11, 2023, Peebles completed his Berkeley PhD dissertation, “Generative Models of Images and Neural Networks,” advised by Alexei Efros. The dissertation covered DiT and diffusion models trained on neural network checkpoints that can sample parameter updates achieving a desired metric. He then joined OpenAI, where, by his own account, he worked on Sora from his first day.⁠[5][2]

On February 15, 2024, OpenAI unveiled the video-generation model Sora with the technical report “Video generation models as world simulators,” of which Peebles was a co-author. That month he was described as co-lead of Sora at OpenAI.⁠[6][3]

On September 30, 2025, OpenAI released Sora 2Sora (OpenAI)Sora is a text-to-video model and app of the same name developed by OpenAI; it debuted as a research preview in February 2024, was followed by Sora 2 and a social video app in 2025, and its web and app services were discontinued in April 2026.Open the full entry and the Sora social video app, with Peebles as head of the Sora team. In October he announced on X that the Sora app had passed one million downloads in under five days.⁠[7][2][8]

On April 17, 2026, weeks after OpenAI announced it would discontinue Sora, Peebles announced that he was leaving OpenAI. In his departure post he credited Sora with sparking a huge amount of investment in video across the industry and argued that research like Sora needed room away from the company’s main roadmap, writing, “Cultivating entropy is the only way for a research lab to thrive long-term.”⁠[9][10]

In September 2026 it became public that Peebles had co-founded, with DreamWorks co-founder Jeffrey Katzenberg and former Dropbox CFO Sujay JaswaSujay JaswaSujay Jaswa is an investor and former chief financial officer of Dropbox who co-founded the technology and media holding and investment company WndrCo with Jeffrey Katzenberg and in 2026 co-founded an AI video startup for filmmakers.Open the full entry, a startup training AI video models purpose-built for filmmakers. Its name, funding size and lead investors were not disclosed.⁠[11]

In October 2026 the startup Peebles co-founded was reportedly in talks with OpenAI to gain access to Sora, customize the model and sell software and services on top of it. Whether it would receive the model weights was unclear, but if a deal went ahead the startup would still get deep enough access to substantially customize the model.⁠[12][13]

Personal website: Public link: https://www.wpeebles.com/ (opens in a new window)⁠[1]

X: Public account: https://x.com/billpeeb (opens in a new window)⁠[1]

OpenAI: Official website: https://openai.com/ (opens in a new window)⁠[6]