Why Filmmakers Need Tailored AI Video Tools
Jeffrey Katzenberg, former DreamWorks Animation CEO, is teaming up with Bill Peebles, ex-head of OpenAI’s Sora video model, and Sujay Jaswa, former Dropbox CFO, to create a new startup focused on training AI video models specifically for filmmakers. The venture, reported by The Information on September 9, 2026, aims to build tools that help directors and studios generate high-quality video content using artificial intelligence. The trio brings together Hollywood storytelling expertise, cutting-edge AI research, and operational scaling experience.
Breaking news
Town AI Secures $108 Million Series B, Valued at $1 Billion After $55 M Pre‑Series
Apple Prepares for Major Product Launch Event on October 28
Apple Releases Release Candidates for watchOS, tvOS and visionOS 27.2
Apple Limits iPhone Duo and iPhone 18 Pro Max to Reservation-Only Sales in Japan and South KoreaThe startup will concentrate on developing AI systems that understand cinematic language, shot composition, and narrative flow, addressing a gap in current generative video tools that often lack nuance for professional use. Unlike broad models trained on internet data, this effort will prioritize ethical data sourcing and collaboration with creative professionals to ensure outputs meet industry standards. Peebles’ work on Sora, which demonstrated advanced video generation capabilities, provides a technical foundation, while Katzenberg’s industry connections and Jaswa’s financial leadership aim to support sustainable growth.
How Will This Startup Differ From Existing AI Video Platforms?
Current AI video generators often produce visually impressive but structurally inconsistent results, making them unreliable for professional filmmaking where continuity and directorial intent are critical. Filmmakers require models that can interpret shot lists, maintain character consistency across scenes, and adapt to specific visual styles or genres. By focusing exclusively on the needs of creators, the startup hopes to build tools that integrate smoothly into existing production workflows rather than disrupt them. Early discussions suggest potential partnerships with studios and independent creators to co-develop features that solve real-world problems in pre-visualization, effects prototyping, and content localization.
Unlike general-purpose AI video tools that serve marketers or social media creators, this venture will target the high-barrier, high-expectation environment of film and television production. It plans to emphasize model controllability, allowing users to adjust lighting, camera angles, and performance details with precision. The team also intends to advocate for transparent training data practices, potentially licensing footage from archives or creating synthetic data under clear usage rights. This approach could address growing concerns about copyright and consent in AI training, positioning the startup as a responsible innovator in a space increasingly scrutinized for ethical lapses.
What stage is the startup currently in? The company is in early formation, with core team members confirmed and initial funding discussions underway, though no product or public demo has been revealed yet.
Frequently Asked Questions
Will the startup release its AI models publicly? Details are not available, but the focus on filmmaker-specific tools suggests a possible business model involving licensed software or studio partnerships rather than open-source distribution.
How will the startup handle data rights for training? The team has indicated plans to prioritize ethical data sourcing, potentially using licensed content or professionally generated synthetic data to avoid copyright issues, though specific policies have not been disclosed.
