Tribeca films show AI works best as a helper, not a whole movie maker

In short: Tribeca 2026 screenings suggested AI in filmmaking works better as a custom tool guided by artists, not as a system that makes a full film from text prompts.

What's going on

Generative AI, which is software that makes images or video from instructions (like asking an app to “make a rainy street at night”), has been widely discussed as a future Hollywood tool. But The Verge reports that most AI video systems still struggle to produce longer, consistent scenes. They often create short clips where details change from shot to shot, like a character’s face or the look of a room.

Tribeca showed both the limits and the more practical uses. Some films leaned heavily on AI and felt rough, including the animated short Roar and ChikaBOOM!, which The Verge describes as lacking polish.

Other projects used AI more like a specialized assistant. Google DeepMind’s Dear Upstairs Neighbors, directed by Pixar veteran Connie Qin He, used custom versions of Google’s Veo and Imagen models. These models were trained on concept art made by production designer Yingzong Xin, so the AI could keep the same painted style across scenes. The team still planned scenes with traditional animation tools first, then used the AI to help “render” them into more finished-looking shots.

OpenAI also had films at Tribeca, including Smoked and Mauvais Soleil. This came despite OpenAI previously shutting down its Sora video tool, which affected other planned releases.

What to watch

The bigger shift to watch is whether studios start building custom AI systems for specific projects, instead of relying on one-size-fits-all tools. That approach is more like tailoring a suit than buying one off the rack, and it may be the difference between AI that supports artists and AI that produces inconsistent footage.

Source: The Verge AI