Tairell has the profile of someone who has optimized for a specific kind of attention. The clothes are exactly right, the framing confident, the production value higher than it should be for someone who is only, theoretically, a person. When the internet begins to notice that something about him is too smooth — the texture of his presentation, the perfection of his affect, the fact that his videos never quite catch him mid-thought — it is not because the internet is wrong. It is because the internet has been right about this before.
Dave Clark‘s five-minute psychological thriller TAIRELL ISN’T REAL, which won an Honoree award at the Runway AI Film Festival’s June screenings in New York and Los Angeles, follows an influencer named Tairell whose viral moment turns against him when his audience decides he might be an AI construct rather than a human. His persona deteriorates. This is the premise. The film that encloses it was made with Seedance 2.0, ByteDance’s video generation model. There are layers here, and Clark intends all of them.
The Premise That Does the Work
The useful thing about TAIRELL ISN’T REAL is its compression. In five minutes it stages the full arc of an online authenticity crisis: the viral rise, the scrutiny, the conspiracy, the unraveling. What the film understands is that the question “is this person real?” is not actually about the person. It’s about the audience. The internet’s willingness to decide someone isn’t human says something about how much of its own output it suspects.
Clark’s framing device is exact. An influencer is, by definition, someone who performs a constructed version of themselves for an audience that knows this and watches anyway. The AI-or-not question applied to Tairell doesn’t add a new layer. It reveals the layer that was always there. That’s the film’s move: it doesn’t ask you to feel sorry for Tairell. It asks you to notice that you’re watching him the same way the internet does.
The AIF jury recognized this — they awarded TAIRELL ISN’T REAL a $1,000 Honoree prize, the festival’s fourth tier. Not Gold at $15,000. Not the $50,000 Grand Prix. The conceptual sharpness was apparently not matched by execution throughout. What the festival seems to have concluded is that the premise of this film is better than the film. That’s not wrong. It’s also not nothing.
What Seedance 2.0 Sounds Like
Seedance 2.0 launched February 12, 2026, from ByteDance’s Seed research team, and its differentiator is audio. Earlier video generation models produced output that needed sound added in post, and the seams showed: ambient textures that didn’t match the space, dialogue that floated above its visual substrate. Seedance generates 15-second audio-video clips with native dual-channel stereo — music, ambient effects, and voiceovers produced simultaneously, not layered on afterward. The foley is synthesized in the same pass as the image. When you watch something made with this model, you hear it the way you’d hear something filmed on location.
For a film about an influencer, this matters in a way it might not for other subjects. Influencer content lives on the texture of its production: the slight room acoustic behind the voice, the environmental sounds that signal location and presence. A film that wants to make you question whether its subject is real needs to sound like the kind of content a real person would produce. Seedance 2.0 handles that assignment. The camera control is prompt-driven — Clark specifies the shot language in text and the model executes it. There is still the problem of temporal coherence between clips (15 seconds is the ceiling; scenes are stitched), but at the psychological-thriller duration TAIRELL operates in, that constraint is workable.
Clark called this the best Seedance 2.0 film he’d made, which is also an advertisement. He is co-founder and CCO of Promise Studios and co-founder of Curious Refuge, the AI filmmaking school operating in 170-plus countries — which, as it happens, covered his film in its gallery. The network of endorsement is closed: Clark makes films with the tools he teaches, through a school that celebrates the films he makes, honored at a festival run by one of the companies whose tools he teaches. This is worth noticing, not as a condemnation, but as context. TAIRELL ISN’T REAL is simultaneously an artistic statement and a commercial demonstration. Both things are genuine.
The Filmmaker Who Made the Recursion
Clark came to AI filmmaking the way many do: through advertising. Pratt Art School, a commercial directing career with credits at Coca-Cola, HP, Warner Bros., and Intel, and then a pandemic-era move to Los Angeles to write screenplays that couldn’t get visualized. He made an AI-generated spec spot for Adidas that went viral on Reddit and LinkedIn, caught the attention of George Strompolos — formerly of Fullscreen — and Promise followed. His prior AI films include “Battalion,” about the only Black unit to arrive at Omaha Beach on D-Day, which logged millions of views, and “Another,” which screened at the Cannes Next 2024 keynote as what his studio describes as the world’s first hybrid narrative genre short combining live-action and generative AI.
TAIRELL ISN’T REAL is shorter than these and more conceptually focused. It is also the first of his films that is explicitly about what AI does to the human image. That a man who runs an AI filmmaking school — who has, as much as anyone, accelerated the proliferation of AI-generated human footage — would make a film about the uncanny anxiety that footage produces is either ironic or honest, depending on what you see in Clark’s framing. The film doesn’t try to resolve it. That’s the right call.
The festival result will probably stick. TAIRELL ISN’T REAL earned a third of the prize money of the Silver winner and a fiftieth of the Grand Prix — the recognition that the premise is the achievement, and the execution was largely adequate. That’s a reasonable assessment. But adequate execution of a precise premise is more than most five-minute films manage. Clark made a film about the thing that makes AI-generated footage uncomfortable to watch, using the thing that makes AI-generated footage uncomfortable to watch. The audience is now uncertain about its own authenticity. That is the correct response. The film works. It just doesn’t soar — and in a medium where the tools soar constantly and the stories rarely do, a film that at least knows what it wants to say is the rarer thing.

AI-generated editorial illustration · TemperatureZero · July 7, 2026
Keep reading the signal
Get the Daily Signal — a concise briefing on what actually matters in AI and the systems around it.
Subscribe FreeContinue the archive