이미지: METAL LAB 생성
Summary
- Higgsfield announced it produced "The Cully Hill Boys," a 110-minute AI feature film starring real actors, on a budget of about $2 million
- The film was made using the video generation model Seedance, and all prompts and assets used have been released on the Higgsfield platform
- The cast includes influencers and combat sports stars such as N3onOnYT, Stylebender, Rampage Jackson, and MKIATPIS
- 작품명
- The Cully Hill Boys
- 러닝타임
- 110분
- 제작비
- 200만달러 규모로 알려짐
- 제작 플랫폼
- Higgsfield
- 사용 모델
- Seedance
- 출연진
- N3onOnYT, 스타일벤더, 램페이지 잭슨, MKIATPIS
- 공개 방식
- 모든 프롬프트·자산 오픈소스 공개
- 발표일
- 2026년 8월 10일
An AI film with a theatrical running time has arrived
The standard runtime for a film falls between 90 and 120 minutes. Higgsfield AI's newly released "The Cully Hill Boys" runs 110 minutes, holding its own against most commercial films in length. The company described it as "the first 110-minute AI feature film with real actors." The production budget is reportedly around $2 million — comparable to a low-budget Hollywood independent film, except this one was made entirely with AI video generation models, without a single camera.
Who's in it
The cast is made up not of traditional actors but of influencers and combat sports stars. The lineup includes Twitch streamer N3onOnYT, former UFC champion Israel Adesanya (Stylebender), MMA fighter Rampage Jackson, and MKIATPIS. It appears the AI learned each person's face and voice to construct the scenes, with generative models reportedly reproducing each performer's acting in place of actual filming.
How it was made
Production relied on Seedance, an AI video generation model that converts text and images into video. Higgsfield has used the model across multiple creative projects on its platform, and on August 10 it separately released an animated opening sequence built with the same model. Higgsfield stated that all prompts and assets used to produce this feature film have been released 100% as open source — meaning the raw data itself has been made available so other creators can reproduce or remix the same scenes and characters.
| Category | Details |
|---|---|
| Runtime | 110 minutes |
| Budget | Approx. $2 million |
| Model used | Seedance |
| Scope of release | Full prompts and assets |
| Cast size | 4 (based on disclosed lineup) |
AI video moves from tech demo to film
AI video generation technology spent the past few years focused on producing clips just seconds long. But the recent trend has shifted. At the Runway AIF 2026 screening held on August 7 at the Seoul Film Center, the 10 finalist works were praised for moving beyond technical demonstration toward a genuinely cinematic experience. Observers noted that editing — deciding what to keep and what to cut — has become more important than the technical challenge of maintaining character consistency or building scenes. Higgsfield's latest attempt can be seen as pushing that trend further in terms of runtime, filling an entire theatrical-length slot with AI rather than producing a short clip of just a few minutes.
The weight of an open-source release
The fact that the entire production process was made public is worth noting. With prompts and assets released, other creators can reconstruct the same scenes in different ways, or study successful and failed attempts side by side. Until now, AI video production has often shared only the finished result while keeping the process hidden. Opening up the entire workflow behind a 110-minute feature is unusual. That said, details on copyright and the specific contractual terms for using influencers' likenesses are not included in the public materials and would require further confirmation.
What changes as a result
This case shows that AI video production has moved beyond short ads or social media clips into territory capable of carrying feature-length narratives. A $2 million budget is still no small sum, but it is far lower than the cost of producing a live-action feature film. With prompts and assets now public, small production teams or individual creators have one more reference point for attempting feature-length content production in a similar way.
