Maggie Gyllenhaal Ditched AI Experiment That Would’ve Made Dakota Johnson Look Like Marilyn Monroe

Maggie Gyllenhaal tested artificial intelligence for her new short film ‘Flesh Impact’, which imagines Marilyn Monroe living past her death at 36. The commissioners behind the project pushed her to use AI, so the filmmaker decided to give it a shot.

The plan was straightforward: train an AI system on licensed Marilyn Monroe footage, then digitally map Monroe’s face onto Dakota Johnson’s performance.

Speaking at the Venice Film Festival, where she’s serving as jury president, Gyllenhaal explained that Johnson’s live acting simply outshone the computer-generated version. “It fundamentally did not work,” she said, noting that the AI recreation lacked the emotional weight and humanity of Johnson’s actual portrayal.

Gyllenhaal scrapped the AI footage entirely for the final cut.

The 17-minute short instead features Johnson as a younger Monroe and Oscar winner Ellen Burstyn as an older version of the Hollywood icon. Without the digital recreation, the film can focus on exploring Monroe’s identity and legacy through genuine performance.

Gyllenhaal also raised concerns about AI’s environmental footprint during her Venice appearance, pointing to the massive energy costs required to run data centers that power the technology.

Her decision reflects a larger conversation in Hollywood about whether machines can replicate what actors bring to the screen. In this case, at least, the answer was no.