OpenAI backs AI-animated feature film 'Critters' targeting Cannes 2026 debut
Sep 8, 2025
Key Points
- OpenAI is backing a feature-length animated film called Critters targeting a May 2026 Cannes debut, using AI rendering tools paired with human artists, writers, and voice actors to compress typical three-year animation cycles into nine months on a sub-$30 million budget.
- The hybrid production model—where AI handles final rendering while humans retain control of concept, script, and direction—signals OpenAI views generative tools as craft assistants rather than replacements, positioning the film as proof of feasibility for entertainment companies wary of guild backlash.
- If AI production cuts costs by 70 percent, the same budget could fund three films instead of one, replicating the startup proliferation effect of cloud computing rather than cannibalizing industry profits.
Summary
OpenAI is backing a feature-length animated film called Critters, targeting a May 2026 debut at Cannes Film Festival. Chad Nelson, a creative specialist at OpenAI, led the project after sketching characters three years ago while experimenting with DALL-E. Vertigo Films and Native Foreign, production companies based in Los Angeles and London, are producing the film, which aims to compress a typical three-year animation cycle into nine months.
The budget is less than $30 million, a fraction of standard animated feature costs. The production strategy reveals where OpenAI sees AI tools as genuinely useful and where human work still matters. The team hired real artists to create hand-drawn sketches that feed into OpenAI's rendering tools, working at storyboard level and letting AI handle final rendering. Human actors voice the characters, and human writers, including some who worked on Paddington in Peru, wrote the script. This hybrid approach uses generative tools as part of a traditional creative pipeline rather than attempting full end-to-end AI film production.
The team retained human voices deliberately. Animated characters can accommodate visual inconsistencies through stylization, but voices are compared directly to human speech and remain more obviously artificial. This suggests OpenAI either lacks confidence in voice generation at this quality level or believes audiences still expect human performance in that domain.
Nelson argues that a real-world case study carries more weight than internal demos. Entertainment companies skeptical of AI adoption need proof of feasibility before risking guild backlash. Disney, Netflix, and others have experimented with AI in production and marketing but have largely avoided wholesale adoption due to pressure from writers' and actors' unions, which fought for strict AI restrictions in last year's contract negotiations.
If AI production tools lower costs by 70%, the same budget that funds one film could fund three. This doesn't necessarily cannibalize industry profits if total production budgets remain fixed. Instead, it redistributes capital across more projects. The parallel is to cloud computing's effect on startups. Lower barrier to entry led to more attempts, higher failure rates, and steep power-law outcomes, but more shots on goal overall.
Bringing a film to completion still requires human choices on concept, script, voice direction, and editorial vision. The comparison to RenderMan and Houdini is the operative frame. These tools became standard options in a craftsperson's toolkit rather than replacements for craft. The reinforcement learning problem of making a profitable film remains hard and is years away from automation.
Warner Bros. Discovery, Disney, Comcast, and Universal have all sued Midjourney for making copies of their copyrighted properties. The details of those cases remain unclear and untested in court. Whether they involve unauthorized training or unauthorized reproduction during production is still being litigated.