Sources: Jeffrey Katzenberg, ex-OpenAI Sora head Bill Peebles, and ex-Dropbox CFO Sujay Jaswa plan to launch a startup to train AI video models for filmmakers
Hollywood mogul Jeffrey Katzenberg is teaming up with the former head of OpenAI's Sora app to launch a new AI startup that would train …
Context & Ripple Effects
The reported venture sits within a multi-year effort to bring generative video into studio workflows: OpenAI was seeking Hollywood partnerships around Sora in 2024 and was still discussing its creative and commercial use with studios in 2025. Those talks were constrained by studio concerns over training data and labor unions.
Katzenberg's Hollywood position and Peebles's Sora experience would make a filmmaker-focused model company a more specialized entrant than a general-purpose video tool. The plan remains unconfirmed, but it highlights that access to professional creative markets depends on commercial terms as much as model capability.
First-order effects
- If launched as reported, the company would give filmmakers a video-model supplier built around production use cases, while placing a former Sora leader in direct competition for that customer base.
- Studios and talent representatives would gain another prospective counterparty for negotiating model-training access, against the backdrop of earlier content-licensing talks with major AI companies.
Second-order effects
- OpenAI, Alphabet, and Meta would face greater pressure to show that their video products can accommodate filmmakers' training-data and labor concerns, rather than relying only on broad creative-tool adoption.
- For studios that have resisted licensing content, an independent filmmaker-focused entrant increases the value of clearly defined rights, attribution, and compensation terms before providing catalog access.
Third-order effects
- If filmmaker-focused entrants win licensed or otherwise approved training access, AI video competition may separate into general-purpose generators and industry-specific systems organized around rights-cleared inputs and production relationships.
- The central bottleneck for commercial video models would shift toward negotiating with rights holders and labor stakeholders, reinforcing AI content commercialization as a distinct market layer.
The trend: Generative-video companies are moving from showcasing models to competing for rights-cleared content, filmmaker trust, and production-specific distribution.