The researchers behind Stable Diffusion and Meta’s Make-A-Video are gathering for a free live community event on November 30th, offering a rare chance to hear directly from the people building the generative media stack. The lineup is stacked with names that will be familiar to anyone following the space: David Ha, Head of Strategy at Stability AI, will discuss collective intelligence and creative AI. Devi Parikh, Research Director at Meta’s FAIR lab, will present Make-A-Video, the company’s approach to text-to-video generation that notably does not require text-video training data.
Patrick Esser, a Principal Research Scientist at Runway, leads applied research including the core latent diffusion model behind Stable Diffusion. His talk, titled ‘Food for Diffusion,’ suggests a focus on what these models need to improve. Justin Pinkney from Lambda Labs will cover extending Stable Diffusion beyond text prompts — he’s the person behind Toonify, Stable Diffusion Image Variations, and Text-to-Pokemon, so expect practical techniques rather than theory. Apolinário Passos of Hugging Face rounds out the program with a talk questioning what comes after the generative media hype cycle.
The event is organized as a companion to a new release, though the announcement doesn’t specify what that release is. Registration is open now through a form, and the organizers promise more details on speakers and talks. The format appears to prioritize high-level accessibility — the talks are framed as a presentation of diffusion models and the tools available for building applications with them.
What’s notable here is how much institutional weight has consolidated around diffusion models in under two years. The speaker list spans Stability AI, Meta, Runway, Lambda Labs, and Hugging Face — five organizations that collectively control much of the open-source generative AI tooling today. Hearing Esser and Pinkney discuss the same underlying architecture from research and applied perspectives could clarify where the bottlenecks actually are. The registration link is live, and for anyone building on these models, the opportunity cost of skipping it is probably higher than the time commitment.
