AI-resurrected performances: a technical miracle with a human cost
AI actor resurrection is the emerging practice of using generative algorithms to reconstruct or extend performances from archived recordings so deceased actors can appear in new or unfinished films, raising creative possibilities and ethical concerns about authenticity, consent, and the future of acting. The technology has moved beyond thought experiment into production reality. AI tools now let creative teams mine old takes, voice sessions, and abandoned edits, then generate missing pieces with eerie continuity. That promise thrills technologists and some producers: nothing is ever truly shelved again, no performance ever definitively over. Yet the same tools provoke unease among filmmakers who see cinema as a record of fragile, unrepeatable human moments. For them, the push toward deepfake film production is not innovation but a philosophical gamble about what a performance is—and whether it should outlive the person who gave it.
American Haiku’s John Hurt experiment shows what AI can unlock
The most striking example so far is Awful Dreams, a short horror film reconstructed around Sir John Hurt’s voice. In 2010, Hurt recorded narration for a project that collapsed after years of funding setbacks, leaving his performance unheard and the film unfinished. Nearly 16 years later, American Haiku founder and CCO Thom Glover revived it with director and AI artist Michael Hess, using generative AI to finally bring Awful Dreams to life. Instead of chasing seamless realism, the team embraced AI’s distortions and hallucinations to create a nightmarish visual language that would have been impossible when the film was conceived. That choice matters: it treats the machine not as an invisible stand-in for the actor, but as a visible collaborator that reframes his work. The project has already found significant success on the festival circuit, where its unusual blend of archival performance and emerging technology has drawn attention.
James Gray’s revolt: cinema as imperfection, not code
While some celebrate AI’s ability to rescue lost projects, James Gray sees the broader AI wave in filmmaking as an aesthetic and moral dead end. He has called the current push for AI-driven filmmaking “obscene,” and says he increasingly believes it will ultimately fail. For Gray, cinema’s power lies in what machines are worst at: “the infinite layers of the soul,” and the small imperfections that make an image feel haunted rather than processed. His response has been almost countercultural. On his new film Paper Tiger, he and cinematographer Joaquín Baca-Asay returned to celluloid, vintage lenses, and tungsten lights as a way to restore “humanity” to the frame. Paper Tiger is set to open this year’s New York Film Festival, before a theatrical release on November 13, 2026. Gray compares this analog resurgence to vinyl’s comeback, arguing that people can hear—and see—the warmth that zeros and ones flatten.
The ethics we cannot outsource to algorithms
Awful Dreams and Paper Tiger capture a split that will shape the next decade of deepfake film production. One camp sees AI as a tool to reanimate stalled projects and expand what images can do; the other fears it will erode the melancholy of film’s “irretrievability,” the sense that each captured moment will never return. Gray warns that digital culture already treats images as disposable, a problem that AI-generated footage threatens to amplify. American Haiku, by contrast, argues that the question is less what AI replaces and more what it enables, especially when an older performance meets a production process that only recently became possible. The unresolved issue is not technical but ethical: even when the result is moving, who decides when a dead actor works again, and on what terms? Until filmmakers confront that, every AI resurrection will carry a ghostly aftertaste.
