AI in filmmaking usually turns into a scandal. This week it turned into a straight answer instead. Danny Boyle confirmed that his new film Ink, which opened the Venice Film Festival before playing Telluride, uses about thirty seconds of generative AI: a couple of old photographs brought to life, and some digital mice running around a newsroom set. He said it on camera, without the usual dodge directors give when this question comes up.
What Boyle actually used AI for
Ink tells the story of Rupert Murdoch’s rise at The Sun in the 1970s, and it needed to bring archival-style images of real figures, including the Kray twins and Margaret Thatcher, briefly to life on screen. Boyle used generative AI to animate a handful of old photographs for that purpose, and separately to generate the movement of some mice in the newsroom. Thirty seconds total, in a feature-length film.
That is a VFX decision, not a performance decision. Nobody’s face was replaced. No actor’s voice was cloned. It sits closer to a matte painting or a crowd duplication shot than to anything that touches a performer’s work.
Why an honest answer is the real story
Social media had already started speculating about the Kray twins and Thatcher footage before Boyle said anything. That is the normal pattern now: a festival premiere, some sharp-eyed viewers, a wave of speculation, and then a studio either denies everything or goes quiet. Boyle skipped that entire cycle and just described what he did and why.
Coming from a director with an Oscar and a filmography that includes Trainspotting and 28 Days Later, that kind of directness carries weight. It reframes the AI conversation in film from a yes or no question into a question of degree and disclosure, which is a much more useful place for the industry to actually be.
The line Boyle draws
Boyle was careful to separate two very different uses of the technology. He called AI a very useful tool in many ways, the same debate this site covered from the writing side, but he drew a hard line at replacing actors, saying that performers should be paid if an AI version of them is used in their place.
That is the distinction that keeps getting lost. Using AI to animate a still photograph for three seconds is not the same category of decision as generating a synthetic performance. One is a visual effect. The other replaces the thing actors are actually paid for.
What this means if you direct
For anyone directing on a limited budget, the practical takeaway is not “use more AI” or “avoid it entirely.” It is that the tool should sit in the same category as any other VFX or compositing choice: it fixes a specific, small problem you cannot otherwise afford, and it never touches a credited performance.
A student film that cannot afford a real 1970s newsroom full of extras and props might reasonably use AI the way Boyle did, for background texture. A film that generates a lead performance to avoid paying or directing an actor is doing something categorically different, no matter how the shot is described afterward.
A simple rule for your own set
Ask what the shot would have needed without AI. If the honest answer is a background detail, a texture, or an object, a VFX-style use of AI is defensible and worth being upfront about. If the honest answer is an actor, the tool is not solving a budget problem, it is replacing a person, and that changes what you owe them.
The takeaway
The mice and the photographs are not the point. The point is the model Boyle just demonstrated: name exactly what AI touched, keep it out of anything a performer is credited for, and pay actors if a synthetic version of them ever stands in for the real thing. That is the test worth applying to every AI story to come out of a film set, not whether AI was used at all, but whether it was used honestly and kept away from the performance.
This article is part of the ongoing craft discussion at Frame One. For a full method on developing a first screenplay from the ground up, see The Screenwriting Series.
Leave a Reply