News

An underwater video generated by the AI model Marey
 

As photographers and filmmakers, many of us are observing the rapid emergence of generative AI tools with a mixture of skepticism and trepidation. We’re doubtful that AI can replace us—especially as we’re working in the underwater realm—but we’re alarmed at the idea that our content is being “scraped” and at some point, AI software will indeed be able to generate the images and video we work so hard to capture. Which makes the arrival of Moonvalley’s AI model Marey both encouraging and scary—not least because the company shows off impressive underwater images and video generated by the model.

While AI company Midjourney is being sued by Disney and Universal for copyright infringement, Moonvalley—which was founded by DeepMind researchers and has close associations with the film industry—is developing Marey by training it on licensed data and with the consent of filmmakers. This should mean filmmakers and studios can avoid the ethical quagmire and copyright lawsuits that have become all too common in the nascent AI industry. Naeem Talukdar, Moonvalley’s CEO and co-founder, tells TIME: “We have to make sure that we’re building these tools the right way: building with the filmmaker and the artist at the center of it, rather than trying to automate their job away.”
 

An underwater image generated by the AI model Marey
 

Aimed at pro filmmakers, Marey offers functionality that many other AI-powered video generation tools lack. Most AI video models are black-box systems: They generate a scene from your text prompt, and if you try to modify one aspect, others may change, making it difficult or impossible to achieve the result you want. Marey aims to offer filmmakers precise control over every detail. You can input storyboards or frames and then tweak the results to taste. You can create cinematic camera moves using just a single image by turning a 2D scene into a 3D environment. You can draw a trajectory for an element and watch your direction come to life. You can even pull motion from a reference video and apply it to new subjects or scenes.

“It’s this iterative process where you start with some input guidance and then you build up towards the scene that you want, which really isn't very different from how VFX workflows are today,” Talukdar tells TIME. “If you’re an independent studio that doesn't necessarily have massive infrastructure, you can now, even in a small space, create and curate these scenes in a very granular way.”

Moonvalley claims that Marey is trained entirely on footage licensed from intellectual property owners. According to Talukdar, around 80% of that footage is B-roll created by independent filmmakers and agencies, and this means the model is trained on about one-fifth of the data used by competitors like Google’s Veo 3. Talukdar says Marey would definitely be more powerful if they scraped data, but he claims they are overcoming this with better technology. “Our inclination is that you don’t necessarily have to be the number one model—you just need to be among the best,” he says.

Marey is now available to filmmakers for subscription tiers of $15, $35, and $150 a month.