After viewing writer/director Eliza McNitt’s new short film AncestraI can understand why Hollywood studios are interested generative AI. Many of the shots in this film were created and refined with only prompts in collaboration with Google DeepMind. It’s clear what Darren Aronofsky and his AI-focused Primordial Soup production housestand to gain from this kind of creative work flow. When you listen to McNitt’s and Aronofsky’s discussion about how the short was created, it’s hard not to think of generative AI and its potential to usher a new era in “content” that feels like it has been cooked up in a laboratory — and put scores filmmakers out to work in the process.
Inspired from the story of McNitt’s own complicated birth Ancestra zooms into the life of a pregnant mother (Audrey Corsa), as she prays that her soon-to be-born baby’s heart defect will miraculously heal. The short film features real actors on practical sets. However, Google’s Gemini Imagen and Veo models are used to create the shots of the mother’s thoughts and the tiny hole in the baby’s chest. We see the baby’s heartbeat in the mother’s womb. The film’s soundtrack is gradually accompanied by Blonde -esque close ups. And the woman’s ruminations on what it means to be a mother are visualized as a series of very short clips of other women with children, volcanic explosions, and stars being born after the Big Bang — all of which have a very stock-footage-by-way-of-gen-AI feel to them.
The message about the power a mother’s loving is cliched. Especially when it is juxtaposed to what is essentially computer-generated nature footage. Ancestrais a project that tries to prove the AI slop videos on the internet are not to be taken seriously. The film’s lack of narrative substance makes it feel like a weak argument for Hollywood’s rush into the slop trough while it’s still hot.
When McNitt smash-cuts to quick shots of different types of animals nursing their young and close ups of holes being stuffed by microscopic creatures, you can tell these visuals are a big part of the film’s AI. Each of these videos is a good example of how text-to-video models can produce uncanny, decontextualized video that would be hard to integrate into a fully produced film. In the behind-the scenes making-of video Google shared in its announcement last week, McNitt spoke at length about why, faced with the daunting prospect of having a real child cast, it made more sense to her that she create a fake baby with Google’s models. McNitt says that there’s nothing like the emotion and performance of a real actor. “But when I was writing that there would be an infant, I didn’t know how we would [shoot] because you can’t[shoot]a baby act.”
The production of films with babies presents a number of challenges that are not present with CGIor doll props. McNitt also had the chance to personalize her film by using old photographs of herself as an infant to create the fake baby’s facial features.
By fine-tuning the shots of Corsa, Ancestra’sproduction team was able create scenes where they almost, but were not quite, interact as if both actors were real. You can see the mother’s hand hovering over her child in wide shots because the baby isn’t there. The scene is so fast-paced that it doesn’t stand out. It’s also less “AI” than the film’s more imaginative shots meant to represent how the mother’s wish heals the baby’s hole in the heart.
Although McNitt notes that “hundreds” of people were involved in creating Ancestra one of the biggest takeaways from the behind-the scenes video is how small the production team was, compared to a more conventional short film telling the story. Hiring more artists to conceptualize Ancestra’svisuals would have made the film more time-consuming and expensive to finish. These are challenges that are difficult to overcome, especially for indie filmmakers or up-and-coming creators who do not have unlimited resources.
Image: Google.
But Ancestrafeels like a case-study on how generative AI could eliminate jobs that were once held by people. AI is often referred to as a tool and jobs will be shifted rather than replaced. It’s hard to believe that studio executives would genuinely believe in a future when VFX specialists and concept artists will be able to transition into jobs as prompt authors who are paid well enough to support their livelihoods. Hollywood’s TV and film actors and writers went on strike in 2023 because of this. It’s why video game actors have been on strike since the beginning of last year. It feels irresponsible dismissing these concerns by saying that people are simply afraid of change or innovation. In the making-of-video, Aronofsky explains that technology has played a major role in filmmaking since its beginning. It’s hard to find a film or TV series today that wasn’t created with powerful digital tools. These tools didn’t exist just a few decades back. The use of generative AI by Ancestramakes it seem like an example of how Google’s models, with enough high-quality data for training, could theoretically become sophisticated enough to produce footage that people actually want to see in a movie theater. Aronofsky’s stony-faced response to DeepMind researcher’s explanation that Veo could only generate eight-second clips is a good indication of where generative AI stands right now, and how is a creative endeavor.
McNitt seems to be revealing a little when she says that the generative models’ output inspired the way she wrote Ancestra (19459023). She says that “both things really influenced each other,” which sounds like a positive spin on the fact that Veo technical limitations forced her to write dialogue to match a series clips vaguely related to the concept of motherhood and birth. It seems that McNitt’s original authorial intention was sometimes sacrificed to work with what the AI models produced. If it had been the other direction, Ancestrawould have told a more interesting story. There’s not much about Ancestra’s narrative, or, to be honest with you, its visuals, that’s so groundbreaking. It doesn’t feel like an example for why Hollywood should be rushing to embrace the technology whole cloth.
While films produced with more generative AI may be cheaper and quicker to make, the technology as it stands now does not seem to be capable of producing art which would push people to sign-up for another streaming service or put butts into movie theaters. It’s important to remember that Ancestra is just an ad to generate hype for Google. This is not something we should rush to do.
