It's almost impossible to browse the Internet without the thrust of AI-generated video. Basically, open a social media platform. It's not long until a creepy-looking clip of a fake natural disaster or an animal doing something impossible slides onto the screen. Most videos look absolutely awful. But they mostly come with hundreds, if not thousands, likes and comments from people who argue that AI-generated content is a new art form that changes the world.
That's especially true for AI clips that are intended to look realistic. No matter how bizarre and aesthetically contradictory the footage is, there are usually people who declare that it is something the entertainment industry should be afraid of. The idea that AI-generated videos are both the future of filmmaking and existential threats to Hollywood has been caught up in a wildfire among relatively new technology boosters.
Often, I find the idea of major studios accepting this technology suspicious given that the output of AI models is not something that could be made into high-quality films or series. This gives the impression that filmmaker Bryn Moother wants to change with the upcoming AI-created feature films from Asteria, the new producer that was launched last year, and Natasha Lyon. He is also an advisor to Late Night Lab, a studio focused on the generator AI acquired by Moother's film and TV film XTR.
A major selling point for Asteria means that unlike most other AI costumes, the generative model built at research company Moonvalley is “ethical” and is trained only with well-licensed materials. In particular, in the wake of the mid-journey appeal for Disney and Universal's copyright infringement, the concept of ethical generation AI could become an important part of the way AI is adopted more widely across the entertainment industry. However, during a recent chat, Mooser emphasizes to me that the company has a clear understanding of what generative AI is and what it lacks.
“When I started thinking about building Asteria, it was clear that as a filmmaker there was a major problem with the way AI was presented to Hollywood,” says Mooser. “It was clear that this tool wasn't built by anyone who'd made movies before. The text-to-video form factor was something that Silicon Valley thought people wanted and actually thought. ”
In Mooser's view, part of the reason why some enthusiasts call the generated video model a threat to traditional cinema workflows is summed up to people, assuming that footage created from the prompts can be reproduced as effectively as what they saw in mimicking and AI-generated music. It was easy for people to replicate the singer's voices with generative AI and create passable songs. But Mooser believes that in a hurry to normalize Gen AI, the tech industry has confused audio and visual output in ways that are in conflict with what actually makes good movies.
“You can't go to Christopher Nolan.” Using this tool, Odyssey“There were some really clear things that were clear as Hollywood people now have access to these tools. One is that the form factor doesn't work because the amount of control filmmakers need is often reduced to pixel levels.”
To provide film production partners with more granular control, Asteria uses the core generative model, Marey, to create new project-specific models trained with the original visual material. This allows artists to build models, for example, that allow them to generate different assets in different styles, and use them to live in a world filled with a variety of characters and objects that adhere to a unique aesthetic. It was the workflow Astalia, which is used to create musician CUCO's animated short, “Love Letter to LA.” By training Asteria's models with 60 original illustrations drawn by artist Paul Flores, the studio can generate new 2D assets and convert them into 3D models used to build the fictional town of video. The shorts are impressive, but their heavy stylisation speaks to how projects using AI generated in the core must work within the visual limits of technology. I don't feel that this workflow still offers control up to pixel level.
Moother says that depending on the financial arrangement between Astalia and its clients, filmmakers can retain partial ownership of the model once the model is completed. In addition to the original license fee, Asteria pays the creator of the material that its core model is trained to. The studio is “exploring” the possibilities of revenue sharing systems. For now, however, Mooser is focused on acquiring artists, with the promise of lower initial development and production costs.
“If you're doing an animated Pixar film, you might come as a director or writer, but there's not much ownership of what the studio makes, the rest, or the cut when you sell a lunchbox,” Mooser tells me. “But if we can use this technology to reduce costs and allow us to fund independently, there is a world where we can have a new funding model that allows for real ownership.”
Astalia will test many of Mooser's beliefs in the transformational potential of generating AI. The eerie valleyA feature film co-written and directed by Lyonne. The live-action film focuses on teenage girls whose unstable perception of reality makes the world more viewed as a video game. Many The eerie valleyFantastic, matrix– Something like visual elements will be created using Asteria's in-house model. I'll make the details in particular The eerie valley It sounds like a project designed to present hallucinatory contradictions that have become known as clever aesthetic features rather than bugs. But Moother says he hopes that “no one thinks about that part” because “everything will touch the director's human.”
“It's not that you're just texting, they're going into video games and seeing what happens, because no one wants to see that,” Moother says. “It was very clear because we were thinking about this. I don't think anyone wants to see what a computer dreams of.”
Like many generative AI advocates, Mooser sees the technology as a “democratization” tool that makes art creation more accessible. He also emphasizes that under the right circumstances, generative AI could easily make films from around $10 million instead of $150 million. Still, securing such capital is a challenge for most young, up-and-coming filmmakers.
One of the big selling points of Asteria that Mooser has repeatedly mentioned to me is the generative potential of AI to generate finished pieces faster and smaller teams. He framed the aspects of the AI production workflow as a positive that allows authors and directors to work closer together without having to spend more time with key collaborators such as ART and VFX supervisors. But, by definition, small teams are converted to less work, raising the question of the possibility that AI can remove people from work. When I bring this up with Moozer, he points to the recent closure of the VFX House Technicolor Group as an example of the ongoing upheaval of the entertainment industry, where generative AI hype began to unemployed before the current heat pitch was on.
Mooser was careful not to underestimate these concerns about generative AI as a big part of what plunged Hollywood into a double strike in 2023. However, he is determined to his belief that many industry workers can pivot sideways into new careers built around generating AI if they open up to adopting technology.
“I want to lean towards this moment just as filmmakers and VFX artists were adaptable and people were able to switch from film editing to enthusiastic editing,” says Mooser. “People who are real engineers, such as art directors, cinematographers, writers, directors, actors, etc. have the opportunity for this technology. What really matters is, as an industry, what's good about this, what's bad about it, it helps us to tell the story, and it's actually dangerous.”
What appears to be rather dangerous about Hollywood's interest in generative AI is not the “death” of larger studio systems, but the possibility of this technology that makes studios easier to operate with fewer real people. It's literally one of Astaria's big selling points, and if that workflow becomes the industry standard, it's hard to imagine today's entertainment workforce expanding in a way that can accommodate the transition to a new career. As for the good things about that, Moother knows the point of the right story. Now he must show that his technique, and all the changes that it involves, will work.
