TORONTO — As Hollywood continues grappling with the legal and ethical landmines of AI, a Toronto-based startup is stepping into the frame with a message: this technology doesn’t have to be the villain.

Moonvalley AI, co-founded by ex-DeepMind scientists Mateusz Malinowski and Mikołaj Bińkowski and helmed by CEO Naeem Talukdar, officially launched its video-generation model Marey this week—aimed squarely at filmmakers, not influencers. And unlike many of the major players in the generative AI space, Moonvalley is banking on a clean, licensed data approach that might actually pass Hollywood’s smell test.

“We’re not trying to replace artists,” Talukdar tells The Hollywood Reporter. “We’re building power tools for them.”

Studios have quietly been using Marey since March to create B-roll, background plates, and previs sequences. But as of Tuesday, Moonvalley opened its platform to broader use, debuting new directing tools that let users tweak character poses, move a virtual camera, and even shift set pieces—delivering granular control far beyond the prompt-only interfaces of OpenAI’s Sora or Google’s Veo.

Talukdar, a former filmmaker turned technologist, is blunt in his critique: “Those tools are cool for TikTok, not for directors who reshoot a single shot 150 times just to get the lighting right.”

With Marey, filmmakers can work faster and cheaper without sacrificing control. Studios are even building custom versions of the model trained on footage from individual shows or features, creating bespoke pipelines that could eventually go beyond background elements and into main unit territory.

Moonvalley’s approach also sets it apart legally. Instead of scraping the internet like most of its competitors, the company is cutting licensing deals with filmmakers on platforms like Vimeo and YouTube, as well as small production companies in West Africa, East Asia, and North America. Talukdar says the company pays for 2-3 year training rights but keeps perpetual generation rights—a model he believes will be more sustainable and less legally fraught.

“We’re trying to avoid the copyright crisis altogether,” he says. “Most AI companies have played fast and loose with data, and the entire industry is paying for it.”

The company even runs its own L.A.-based studio, Asteria, where it hires actors and produces original clips to enrich Marey’s training data—and, eventually, to shoot commercial work.

But Moonvalley hasn’t dodged the backlash entirely. Indie icon Natasha Lyonne is using Moonvalley’s tools in her directorial debut Uncanny Valley, co-written with Brit Marling, about a teen obsessed with an AR video game. Though the film aims to comment on tech addiction, its AI-assisted production has drawn heat from film workers and fans who see it as a harbinger of cheaper, lower-quality filmmaking. Talukdar pushes back hard.

“They couldn’t get a studio to greenlight a $30 million budget,” he says. “With our tech, they can make it for a third of that. It’s a new way to greenlight projects that otherwise wouldn’t get made.”

The firm plans to monetize through subscriptions, pricing each generated 5-10 second clip at around $1.50, in line with industry averages.

Backed by $70 million in early-stage capital from Khosla Ventures, General Catalyst, and Bessemer Venture Partners, Moonvalley is now raising a follow-up round as it looks to scale. The company currently employs 50 people, including a dozen in Canada—strategically positioned between Toronto’s booming VFX scene and Vancouver’s global production footprint.

Talukdar knows AI’s entrance into Hollywood isn’t welcome everywhere. Both the WGA and SAG-AFTRA strikes made AI a flashpoint, and VFX artists have openly worried about job displacement. But he insists Moonvalley’s long game is about expansion, not erosion.

“You’re going to see a wave of greenlit projects that would’ve died in development hell,” he says. “This isn’t about replacing humans. It’s about finally giving them the tools they’ve been asking for.”