Hollywood's Future: More Than Just AI Prompts

The Tribeca Film Festival this year presented artificial intelligence in filmmaking with greater nuance than merely "slop." Despite widespread discussion about generative AI's potential to revolutionize the industry, genuinely compelling projects created with the technology, the kind audiences would pay to see, have been largely absent. Most AI firms' video models still only produce short, visually inconsistent footage, and some significant Hollywood AI partnerships have unexpectedly dissolved, suggesting studios may not yet fully rely on Silicon Valley's new technology. Generally, major production houses have seemed limited to creating low-quality, short-form video with generative AI. However, this trend might shift if studios observe the experimental works showcased at Tribeca, which demonstrated how human artists can leverage the technology in compelling ways, even if it feels unlikely that generative AI will ever craft an entire film from scratch.

While none of the AI-powered movies screened at Tribeca were as deficient as some of the "video slop" circulated online by companies like OpenAI and xAI, some projects exemplified why generative content often lacks the vitality of human-crafted art. *Roar*, an animated short by Illuminai Studios, felt more like a disjointed montage of AI-generated clips than a cohesive cinematic piece. Similarly, Asteria Film Co.'s *ChikaBOOM!*, a fast-paced fantasy about a magician in training, lacked the visual and sonic polish necessary to truly captivate an audience.

The overall unrefined quality of *Roar* and *ChikaBOOM!* appeared to reflect the inherent technological limitations embedded in their AI-forward production workflows. In contrast, films like Google DeepMind’s *Dear Upstairs Neighbors* and OpenAI’s *Mauvais Soleil* demonstrated how filmmakers can circumvent these challenges when generative AI is deployed with greater ingenuity.

Co-written and directed by Pixar veteran Connie Qin He in collaboration with Google DeepMind researchers, *Dear Upstairs Neighbors* tells the story of an exhausted young woman desperately trying to sleep. Ada (Márcia Mayer, who also produced the short) simply desires a few hours of peaceful rest before returning to work. Yet, each time she begins to doze off, the cacophony emanating from her upstairs neighbors' apartment jolts her awake, leaving her to ponder their mysterious nocturnal activities.

To give *Dear Upstairs Neighbors* a distinctive world, He enlisted Pixar production designer Yingzong Xin, who created concept art in Photoshop and with acrylics on paper. The expressionistic aesthetic of these illustrations was vital to bringing the film's fantastical story to life, but it also presented a unique hurdle for DeepMind’s researchers. Most AI video generation models would struggle to transform such a painterly style into visually consistent footage. To address this, DeepMind’s engineers developed custom versions of Veo and Imagen, specifically designed to empower *Dear Upstairs Neighbors’* artists with precise control over their outputs.

This approach highlights how filmmakers must actively work around some of generative AI's more typical limitations.

Because the customized models were trained on Xin’s concept art, they could consistently generate shots that adhered to He’s vision for the project. The text-to-video models proved adept at reproducing specific stylistic details, such as the visual representation of sound when objects interact. However, to construct *Dear Upstairs Neighbors’* scenes into a cohesive narrative, the creative team also employed more traditional methods. By creating rough animations with Autodesk Maya, the industry standard for 3D rigging and VFX, the production team ensured that scenes would unfold exactly as desired. These rough animations were then fed into Veo, enabling artists to create more visually polished scenes, ready for further enhancement with additional stylized assets generated by Veo and Imagen.

More than any other film at Tribeca, *Dear Upstairs Neighbors* served as a case study in how generative AI can be utilized as a bespoke tool that genuinely assists artists in developing their ideas. The film's entire workflow relied on human-made art and individuals making the nuanced creative decisions that text-to-video generators cannot achieve autonomously. It's important to recognize that *Dear Upstairs Neighbors* would not have been nearly as visually impressive if it had been produced with standard versions of Google’s various models. While the models performed exceptionally for this particular short, that is also to be expected for a project that concurrently serves as a commercial showcase for Google’s technology.

In comparison to OpenAI’s contributions to this year’s festival, *Dear Upstairs Neighbors* offered a much more enjoyable viewing experience. Alice Gu’s semi-autobiographical drama *Smoked* used Sora to recreate the Palisades Fire, and Youssef Michraf’s *Mauvais Soleil* featured several photorealistic scenes generated with OpenAI’s creative tools. Watching both films, it was evident how their respective filmmakers had to navigate some of generative AI’s common limitations. Wide shots in *Smoked’s* fiery scenes appeared somewhat cartoonish, though the effect worked more effectively in close-ups—filmed using a Volume-like setup—of a woman and her son attempting to escape the blaze in their car. Most of *Mauvais Soleil’s* shots lasted only a few seconds, with an unseen narrator as the sole speaking character; however, the film’s narrative about a man’s life being warped by artificial intelligence made these details feel like intentional artistic choices.

OpenAI’s presence at Tribeca was somewhat surprising given the company’s recent decision to entirely shut down Sora. Sora’s abrupt closure also led to OpenAI’s feature-length film *Critterz* being unable to debut at this year’s Cannes Film Festival. While this suggests OpenAI might be pivoting away from video-focused applications of its technology, other players in the generative AI space continue to develop tools that filmmakers can utilize to realize their projects.

With only $2,000 spent entirely on computing costs, writer/director Ash Koosha single-handedly produced *Dreams of Violets*, a docudrama focusing on the nationwide protests that have swept Iran over the past year. Using Kling AI, Claude, Gemini, and Nano Banana, Koosha tells a fictionalized story about a group of people trapped in an alley as police brutalize civilians in the streets. The project took Koosha just a few weeks to complete by himself, and while it is supported by a powerful narrative, it does not break new visual ground.

After viewing all these films, a distinct impression emerged: there is no foreseeable future where studios will churn out commercially viable projects simply by feeding prompts to generative AI models. While that type of content likely won't disappear, it’s not the kind of material Hollywood’s heavyweights would endorse. A much more probable scenario involves larger AI firms like Google partnering with studios to build bespoke models tailored to very specific workflows. Crucially, these workflows will only function effectively when guided by human artists with very clear creative visions.