Higgsfield has published the production build of The Cully Hill Boys, a 110-minute AI feature film produced for USD 2m in four weeks, under a study-only license that lets anyone read the prompts, character sheets, and a 137-entry production log, but prohibits copying, generative reuse, and derivative works.
The release on August 7, 2026 is the first time a “studio” has published the full AI production build behind a licensed feature film. The accompanying film itself streams free on Higgsfield Originals.
The published materials provide unusually granular documentation of how a voice-and-likeness licensing architecture can convert two-minute voice recordings and photo shoots into a feature-length performance, impacting the studio dubbing industry.
The same architecture produced Kok Boru, a 15-minute Central Asian folktale short with native Kazakh voice performances released by Higgsfield at the SIGGRAPH conference in July 2026.
The two films together show how AI video generation is evolving from a visual-effects novelty into a media localization primitive.
What was released, and where it lives
The Cully Hill Boys’ project canvas exposes the prompts, character sheets, and location assets, plus a 137-item log documenting successive revisions, version history and internal assessments. Sitting alongside the project is a structured prompt system using a CINEDANCE workflow, built for ByteDance’s Seedance model.
The film is hosted on the Higgsfield Originals platform launched in March 2026 with its first title, Arena Zero. The catalogue now includes Hell Grind, Zephyr, Mork, The Cully Hill Boys, and several others, with Kok Boru, Lost in Shibuya, Golden Hour, Peter Pan, and Viking Courier in the release pipeline.
The published materials are a reference project for the company’s Global Film Festival, a USD 1m prize-pool competition that opened on August 10, 2026 and runs through September 3, 2026. Judging is led by five-time Oscar winner Edwin Catmull, former president of Walt Disney Animation Studios, Oscar- and BAFTA-nominated cinematographer Phedon Papamichael and franchise director Paul W. S. Anderson.
The film, in six lines
The Cully Hill Boys is a 110-minute AI-generated action comedy that follows three struggling East London rappers who, while trying to shoot a music video, end up in a chaotic drug war between rival crime syndicates.
The film stars AI likenesses of four licensed performers: two-time UFC middleweight champion Israel Adesanya, MMA veteran and actor Quinton “Rampage” Jackson, livestream creator N3on (Mikyle Rafiq, 21), and streetball athlete Matt “MK” Kiatipis.
The screenplay was written by Tim Planagan over five years and directed by Aitore Zholdaskali and Adilet Abish, who ran a nine-director horizontal production team.
The film premiered at The Glasshouse in New York City on August 5, 2026; however, it is not headed for a theatrical release. Higgsfield has positioned it as a showcase intended for studios, agencies, and brands rather than general audiences.
What USD 2m paid for
In an email exchange with Forbes after the premiere, Higgsfield shared that the total production budget of about USD 2m was split roughly evenly, USD 1m to AI compute and USD 1m to the human team of directors, generative artists, photo and film editors, costume and production designers, and post-production.
Over a four week schedule, a 28-person team produced roughly 1,000 production-ready assets in 4K across more than 100 locations. About 40% of the team had no prior AI experience, with Higgsfield Academy handling their onboarding.
The production did not use an image-to-video workflow. Instead, the team generated shots in Cinema Studio with ByteDance’s Seedance 2.5, using written prompts and reference assets.
The closest comparator in Higgsfield’s own filmography is Hell Grind, a 95-minute action sci-fi produced for USD 500k by a 15-person team in 14 days. The asset-and-locations footprint rose by roughly an order of magnitude between the two releases, with most of the addition going into the larger character cast and the East London locations.
Higgsfield told Forbes that similarly ambitious independent productions used to take one-to-two years and cost more than USD 20m.
The voice-and-likeness licensing architecture
The novel element of the production is not the visual generation but the contractual framework.
The four licensed performers supplied two inputs: two-minute voice recording and high-resolution photographs. The actors are compensated under per-project contracts where they review the screenplay and their character profile before production begins, and source assets are permanently deleted 30 days after production ends.
The economics of that input are what makes the architecture scale. N3on told Variety that the project needed only a two-minute recording and some photos to bring his likeness to life, and that the compensation was “a pretty good check with very little work.”
N3on framed the deal as the only realistic way for a livestream creator to appear in a film: “I’m streaming every day, so me trying to act in a movie and wake up at 6 a.m. and do all these things and come every day, it’s just unrealistic. So this is a way to do it without having to do anything really.”
The deal is the first time a feature-length film has been produced with licensed celebrity likenesses generated entirely by AI. Forbes reported that N3on and Kiatipis’ representatives are adding paid AI-persona licensing to the services they already manage like sponsorships and endorsements.
Alex Mashrabov, CEO and co-founder of Higgsfield, framed the architecture in the statement announcing the premiere: “Building brand equity takes years of hard work for talent, and we are providing creators and studios a powerful new extension of it. By removing physical constraints and reducing production costs, we’re empowering filmmakers, creators, and studios to unlock a whole new talent pool and tell stories that were previously out of reach.”
The localization economics
The same architecture puts the localization industry on a different footing. Higgsfield’s 15-minute animated short Kok Boru was completed in two weeks by a 10-person team spanning animation, writing, sound and color, with native Kazakh actors voicing the characters.
The same workflow that lets a small Kazakh team put Central Asian folklore on a global streaming platform with native-language voice performance is the same workflow that licensed four English-speaking fighters, makers, and creators for a 110-minute feature.
The dubbing studio has, in effect, become a single recorded voice session. Higgsfield’s voice-cloning product can “Clone, Read, Dub & Export Any Voice,” and the Soul Cast feature inside Cinema Studio 2.5 lets creators call saved or newly generated AI actors.
The N3on interview goes one step further: the two-minute voice recording he supplied for The Cully Hill Boys is the same input a voice-cloning product consumes, and the same input a localization team would supply for a dubbed version of any other film.
For language-service and dubbing teams, the regulatory frame is the binding constraint. The 118-day SAG-AFTRA strike produced the 2023 TV/Theatrical Contracts, the first Hollywood agreement requiring the performer’s informed consent and compensation in most cases, with residuals as applicable, before a performer’s digital replica can be used.
The NO FAKES Act of 2026 (S. 4591), now moving through the US Congress, would give “each individual or right holder the right to authorize (1) the use of a digital replica embodying the individual’s voice or visual likeness, or (2) the use of the individual’s voice or visual likeness for a product or service requiring authorization to avoid liability under the bill.”
For language-service and dubbing teams, the public-facing takeaway is narrow but concrete. The output of a typical AI voice-cloning session can now be the input of a feature-length film production. The output of a typical dubbing session can be the input of a streaming-platform release. The line between “voice” and “performance” is now a licensing clause, not a recording booth.
