Hollywood's future with AI lies in custom-built tools guided by human artists, not in feeding prompts to off-the-shelf models, according to films showcased at Tribeca.

What happened: Several AI-powered films debuted at Tribeca Film Festival this year, including Google DeepMind's Dear Upstairs Neighbors, OpenAI's Mauvais Soleil, and Ash Koosha's Dreams of Violets. These projects demonstrated different approaches to using generative AI in filmmaking—from custom model fine-tuning to low-budget independent production with $2,000 spent entirely on computing costs.

Why it matters: Most AI video models today produce short, visually inconsistent footage, and generic AI-generated content lacks the polish and coherence of human-crafted art. The successful Tribeca films showed that filmmakers can avoid these limitations when they combine generative AI with traditional production tools and strong creative direction from human artists—suggesting studios may need to partner with AI firms to build bespoke models rather than relying on vanilla, prompt-based systems.

What to watch: Dear Upstairs Neighbors illustrates the winning workflow: custom versions of Google DeepMind's Veo and Imagen were trained on concept art by a Pixar designer, then combined with rough animations created in Autodesk Maya before final polish. This hands-on, artist-guided approach appears to be the model studios will pursue, rather than fully automated AI generation.