Fal's inaugural Generative Media Conference draws 300 attendees including Jeffrey Katzenberg

Summary

Fal hosted its inaugural Generative Media Conference on October 24th at San Francisco's Ferry Building, drawing approximately 300 attendees including Jeffrey Katzenberg and representatives from major Hollywood production studios, foundation model labs, and enterprise users spanning architecture, gaming, and marketing.

Berkay, speaking from the event, frames the generative media market as structurally distinct from the language model space. Where LLMs have consolidated around a handful of dominant players, generative media is seeing hundreds of companies building, customizing, and deploying models in production, suggesting a more fragmented and competitive landscape.

The conference signals a shift in industry focus from capability benchmarks toward applied workflows. Attendees are less concerned with which model wins on evals and more focused on building application-layer tooling that lets downstream creatives in marketing, film, and architecture actually use the technology. Berkay describes the current moment as a "utility phase."

On Sora 2, Berkay characterizes it as a strong but narrowly tuned model, well-optimized for social-content aesthetics and UGC-style output. He suggests its underlying architecture likely has broader capability, and that OpenAI could in principle fine-tune variants for different use cases, but as shipped, it occupies a specific niche rather than serving as a universal generative video foundation.

Hollywood's posture is shifting. Berkay notes a move from avoidance toward active experimentation, with several projects in the pipeline he expects to surface by late 2025 or early 2026. The industry's reluctance to speak publicly about AI adoption, particularly in Los Angeles where union and creative-labor tensions remain elevated, has not stopped major studios from privately engaging with the technology.