Maggie Gyllenhaal Says She Tried Using AI to Turn Dakota Johnson Into Marilyn Monroe — Then Scrapped It: ‘It Fundamentally Did Not Work’
Venice Film Festival jury president Maggie Gyllenhaal has revealed she tried using AI to digitally transform Dakota Johnson into Marilyn Monroe for her short film "Flesh Impact," which is screening at the festival, but ultimately abandoned the technology because it drained the humanity from Johnson's performance. Speaking at a symposium on the Lido alongside composer Alexandre Desplat and fellow filmmakers, she said the experiment undermined the very reasons the project was conceived, prompting her to scrap it entirely.
Gyllenhaal explained the short depicts a 100-year-old Marilyn Monroe with flashbacks to her 30s, played by Johnson, and that the film's commissioners had asked her to use licensable AI trained on licensed Monroe footage to overlay Monroe's face onto Johnson's. She said the result "fundamentally did not work" compared with Johnson's own performance, and went on to raise wider concerns about AI in filmmaking, including environmental costs from data centres, warnings from "Godfather of AI" Geoffrey Hinton, and scepticism that AI would genuinely "democratise" filmmaking rather than simply enrich those who own the technology, potentially undercutting budgets for traditionally made films.
In short
- Gyllenhaal scrapped AI face-swap of Johnson as Monroe in her short film
- She said AI performance felt less human than Johnson's own acting
- She also flagged AI's environmental costs and threat to film budgets
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Using licensed AI to recreate Marilyn Monroe could have offered a practical way to evoke her recognisable image while preserving a new performance, and might have made a demanding visual effect accessible without extensive conventional effects work. More broadly, supporters argue that AI tools can lower production barriers and widen creative opportunities, provided their use respects rights and is transparent.
The case against
Gyllenhaal’s experiment suggests that replacing an actor’s face can weaken the individual expression and humanity that make a performance compelling, while raising concerns about consent, ownership and the use of a performer’s likeness. Critics also argue that AI’s environmental costs and concentration in the hands of technology owners could outweigh claims of democratisation, while making it harder to sustain traditionally made films.
Explained simply
Marilyn Monroe was a major American film star whose image remains widely recognised. Dakota Johnson is an actor; Maggie Gyllenhaal is an actor and filmmaker. For a short film about Monroe at different ages, Johnson plays Monroe in her thirties.
Artificial intelligence can generate or alter images and video using material it has been trained on. In filmmaking, this can include digitally changing an actor’s face, raising questions about how the change affects a performance and how a person’s likeness is used.
The debate also touches on who benefits from AI tools and what they may mean for film budgets and production. Their computing systems use data centres, which require energy, so their environmental impact is another part of the discussion.
AI Art Culture Entertainment Film Government Politics Research Science Technology