A computer-crafted likeness of the late Val Kilmer appears in the trailer for the western As Deep as the Grave, with the actor’s estate granting permission. The family cited his interest in new storytelling tools, yet the visual replica,rigid and missing his voice,has ignited a debate about whether such technology could eventually make human actors redundant.
Last year the AI-generated “starlet” Tilly Norwood debuted in a comedy sketch, prompting Emily Blunt to call the figure “really, really scary,” according to The Guardian. The synthetic persona is slated to feature in the upcoming feature Misaligned, intensifying concerns that audiences may be presented with performers they never met in person.
Tim Webber, chief creative officer at the British visual-effects house Framestore, doubts that fully synthetic performers can convey genuine inner life. He explains that current models learn only surface details, such as image pixels, and that reaching true emotional depth remains a distant goal. Webber’s skepticism highlights a gap between visual realism and the psychological complexity of human acting.
Higgsfield AI claims to have closed the uncanny valley gap, showcasing two short-form projects. Hell Grind, released in May, limits each 15-second segment to a single emotion using the Chinese video model Seedance 2.0. Three months later The Cully Hill Boys employed the upgraded Seedance 2.5, featuring AI recreations of streetballer Matt Kiatipis, MMA champion Israel Adesanya and livestreamer N3on, demonstrating a rapid rise in expressive capability.
The Cully Hill Boys was assembled by writer Tim Planagan in thirty-second blocks, a process that begins with detailed visual assets of the characters. Each block is driven by a 4,000-word prompt,partly generated by a large language model,that specifies action, lighting, camera moves and the intended emotion. The prompt is refined repeatedly until the output meets the creators’ standards, a labor-intensive workflow that still relies on extensive human direction.
Filmmakers such as London blend real-actor dialogue with retro-fitted animation, noting that the resulting performances suit stylised, larger-than-life roles rather than subtle character work. Tye Sheridan, co-founder of Wonder Dynamics, stresses that AI tools should augment rather than replace actors, citing his experience on Ready Player One and his Flow Studio system that uses AI-enhanced motion capture. Both Sheridan and Framestore’s Webber are exploring “digital makeup” techniques that preserve a human core while leveraging generative technology.