Sora prompted by Walter Woodman, Sidney Leeder and Patrick Cederberg

New studios in Hollywood, a startup valued at billions, and factories for synthetic stars: A whole new industry is emerging around AI-generated videos, complete with investors, business models, and distribution conflicts.

In Culver City, within sight of the historic Sony Studios, the AI studio Promise is currently filming the horror movie “Touch Grass” with a budget in the low millions. As The Guardian reports from a set visit, the soundstage consists of little more than white sheets in a corner of an office.

While a camera films actress Tori Thomas during a chase scene, the AI overlays her movements in real time with a swaying field of grass generated by the Chinese model Seedance 2.5, which was released this month. Even the blades of grass move as the actress appears to brush past them. The entire scene, featuring the actress and the backdrop, appears instantly on the director’s monitor, and with just a few clicks, the meadow can be replaced by an underground cave.

According to The Guardian, Promise is backed by Google, Silicon Valley investors, and Disney, making it part of a growing number of AI studios setting up shop in Hollywood. Co-founder George Strompolos estimates that hybrid productions combining human actors and AI technology are 20 to 50 percent cheaper than conventional films, which could make filmmakers less dependent on the major studios.

A figure from Netflix shows just how widespread the technology already is: The streaming service stated that by 2026, it will have used AI in 300 of its 1,000 titles. Ron Howard is also collaborating with the AI studio Obsidian on an animated documentary, whose executive producer estimates the savings at 30 to 40 percent.

Oscar winner Joel Hynek, who once created the analog effects for “Predator,” compares the upheaval to the shift from film to digital and speaks of “incredible promise.” At the same time, resistance remains significant: Directors such as Christopher Nolan and Guillermo del Toro are critical of generative AI.

Higgsfield Shows How Quickly the Market Is Commercializing

While Hollywood is still debating the use of the technology, the video startup Higgsfield is already profiting from it on a large scale, primarily with promotional videos rather than feature films. According to the Financial Times, the company, which was founded by former Snap executive Alex Mashrabov, has raised $400 million at a valuation of $5.4 billion, with investors including DST Global, Goldman Sachs, Liberty Global, and Intel.

Just eight months earlier, the valuation stood at $1.3 billion, while annualized revenue rose from $20 million to $700 million within a year. Particularly noteworthy is the shift in its customer base: While business customers accounted for less than a quarter of revenue in January, they now make up the majority. According to Mashrabov, brands like Dollar Shave Club are using the platform to produce several marketing videos per day, whereas in the past they would create just a single campaign video.

Synthetic Stars Off the Assembly Line

At the same time, an industry for artificial personalities is emerging. According to the New York Times, the startup Inception Point AI manages over 100 AI personas, each with its own niche, ranging from British gardening expert Nigel Thistledown to DIY influencer Lila Walker. Each character receives a 15- to 30-page “character bible” complete with fictionalized life stories designed to make them seem more relatable.

The business model relies on sheer volume: The company maintains more than 5,000 active podcasts, and because production costs are minimal, CEO Jeanine Wright says that just 20 listeners per episode are enough to turn a profit. Similar trends are emerging in fashion, where the agency Seraphinne Vallora creates AI models for brands like Guess, and in music, where AI avatars like Xania Monet are reaching the Billboard charts. Spotify plans to label such profiles with an “AI Persona” tag in the future.

From Sora’s False Start to a Thriving Industry

Just over two years ago, the market looked completely different. When OpenAI unveiled the first Sora demos in early 2024, the model was far ahead of the competition technically and sent shockwaves through Hollywood. However, the version originally shown never appeared in that form for the public. The version released in late 2024 failed to meet the high expectations, and while OpenAI hesitated, competitors from the U.S. and, above all, from China rapidly caught up. The subsequent launch in the fall of 2025, featuring Sora 2 as a standalone app, failed to translate the initial hype into sustainable success and was eventually discontinued. A billion-dollar deal with Disney fell through.

In contrast, production tools for studios, marketing platforms for businesses, and persona factories for the creator economy have proven to be viable business models. The result is an industry that now generates its own valuations, job profiles, and distribution conflicts.

Read on for the full picture.
Subscribe for hype-free coverage.

• Full access to every article on THE DECODER
• No ads
• Join the comments and community discussions
• A weekly AI news recap via mail
• 6x/year: "AI Radar" — deep dives on the AI topics that matter most
• Daily AI news, always up to date
• Our full ten-year archive
• Covered by a team with 10+ years in AI