Key Takeaways
- Ainos Studio CEO Ham Seul-gi screened “Steelay,” a 90-minute AI-generated feature film, at CGV Yongsan I’Park Mall in Seoul on September 7, 2026, claiming no computer graphics were used.
- The film was assembled from 13,000 AI-generated clips, each roughly 15 seconds long, requiring character and scene consistency across the entire runtime.
- Verifying the “no CG” claim independently is not currently possible with available tools, putting “Steelay” at the centre of a provenance problem the wider industry has not solved. On September 7, 2026, Ainos Studio held a blind screening and press conference for “Steelay” in Seoul, a 90-minute science-fiction film assembled from 13,000 AI-generated clips with no computer graphics, according to its makers. CGV Yongsan I’Park Mall hosted the event, where Ainos Studio CEO Ham Seul-gi presented “Steelay” as evidence that AI can carry a feature-length narrative without the visual effects pipeline that has defined blockbuster filmmaking for three decades.
Building a Feature from 15-Second Shots
Constructing a 90-minute film from 13,000 individually generated clips is the most technically demanding thing about “Steelay.” Each segment ran roughly 15 seconds, which means the team had to maintain character appearance, lighting, environments and plot continuity across thousands of discrete outputs. That is not a post-production problem so much as a generation problem: inconsistency baked in at the clip level compounds badly when you cut 13,000 of them together.
Ham Seul-gi identified consistency as the core difficulty. The production timeline coincided with the availability of more capable generative tools, which the team credits with making the assembly manageable. Whether those tools offered longer clip lengths, tighter parameter control, or more stable character outputs is not specified, but the timing was clearly a factor in getting the project finished.
The “No CG” Claim
Ham has stated directly that “Steelay” contains no computer graphics. Director Seol Woo-in built AI characters by referencing the appearances of existing idols, and actor Kang Tae-kyung is reported to have contributed his face to AI-generated footage, describing the AI as a creative tool rather than a replacement. Ham acknowledged that a direct cost comparison to conventionally produced films was difficult to provide, an honest concession, given that this production pipeline has no clear precedent to measure against.
Who Was Watching
Ha Jung-woo, Vice Chairman of the National AI Strategy Committee, attended the screening and expressed support for AI content’s potential. The audience reportedly had no prior knowledge of how the film was made. That framing positions “Steelay” as more than a technical demonstration: it is being offered as an industry argument.
The conversation in AI video has shifted from “can AI generate film-quality footage” to “can AI sustain a story,” partly because tools like OpenAI‘s Sora have made high-quality generation accessible enough that the first question feels answered. “Steelay” is a direct attempt at the second, and the answer Ham’s team is putting forward is yes, though how the wider industry receives that answer is unresolved. For builders watching this space, sustained autonomous output over long timeframes is the shared challenge whether you are running an agentic workflow or generating a feature film.
Provenance in the Age of Generated Film
A film claiming zero CG and total AI authorship sits at the centre of a verification problem the industry has not solved. Distinguishing AI-generated footage from traditionally recorded material is getting harder as models improve. Visual tells that once flagged synthetic imagery, unnatural motion, inconsistent lighting, impossible textures, are increasingly rare in current-generation outputs.
Cryptographic provenance standards like C2PA and Content Credentials are the most concrete response so far. Both aim to give media files a verifiable history: who produced them, when, and what was changed. By foregrounding its AI-only production publicly, “Steelay” is implicitly participating in that conversation, whether the team framed it that way or not. The harder question, how a viewer or distributor independently verifies the “no CG” claim, is one the film raises without answering. That is not a criticism of the project; it is the live problem facing anyone trying to establish trust in AI-generated content at scale.
Originally published at https://autonainews.com/ainos-studio-unveils-90-minute-ai-film-steelay-claiming-no-cg/
Top comments (0)