A breathing painting
The Poison Taster follows an imperial food taster through the reigns of two emperors in Ancient Rome.
There’s a bear in The Poison Taster, asleep in a hedge it has destroyed. The gardener has spent sixty years working on that hedge. I wanted the scene to hold comedy and melancholy at the same time.
Before this bear there were a lot of other bears. One tearing the hedge apart. One seated in a fountain. One eating roses, one nose to nose with a bronze statue.
I developed the images in Absurd through conversations with agents. As each batch of bears came in, I wrote down what I did and didn’t like about them. Each note pulls the next batch a little closer to what I want, until one of them is the one.
What made the final bear right was how peacefully it slept. It had made itself a comfortable bed out of the gardener’s work. I can feel affection for the bear and sorrow for the gardener at the same time.
A photograph made in Ancient Rome
The visual DNA of this project was built through long conversations about pictures. Early on we talked about street photography, and that decisive fraction of a second in which a scene explains itself. I wanted the images to feel as though a great street photographer had been dropped into the streets of Ancient Rome. We talked about Renaissance and Baroque paintings in the same way: where the light falls, where the people stand, and what those choices can make the viewer feel.
Those conversations helped us translate what I love about the paintings into language we could use in our prompts. A painting could become the starting point for a scene, even if the story it was telling was very different.
For a physician preparing poison beside a handmaid, I drew on Georges de La Tour’s Joseph the Carpenter. The handmaid holds up a candle, and the light shines red through her fingers. The quiet attention of the painting becomes the work of two servants in the middle of the night.
For a bear loose in the emperor’s bedchamber, I used Delacroix’s Death of Sardanapalus: the ruler reclining high on a red couch while everything below him falls into chaos. The scene changes, but the contrast between his stillness and everyone else’s agitation gives us something to build with.
For an imperial banquet, Veronese’s The Wedding at Cana was perfect: a symmetrical court, a gallery above the guests, and abundance spread across the frame. The taster becomes a small figure within it.
Bruegel’s The Tower of Babel gave the amphitheatre under construction its scale: a colossal building seen from above, with workers scattered across it and the emperor dwarfed by his own project.
The prompt for the physician’s study drew on Rembrandt’s philosopher interiors: a seated figure, a winding stair, and window light falling into a dark room.
For a tiger fighting wolves, I used Frans Snyders’s Dogs Fighting. The animals form a tight knot of bodies, with heads and limbs pulling in different directions.
Titian’s The Bacchanal of the Andrians inspired the prompt for a garden revel: wine, music, dancing and reclining beneath the trees.
Places to get lost
The Poison Taster has twenty-five distinct chapters, each with its own page in my Absurd canvas. There are a lot of places to get lost in a project like this. Palaces, kitchens, baths, markets. The list of pages lets me see the shape of my whole film from a bird’s-eye view, even when I’m deep inside a small part of it.
Within each page, the pieces of the chapter are organized in columns and the variations in grids. I can leave a chaotic chase scene half finished and go to work on a much quieter shot, knowing I’ll be able to find my way back.
The canvas remembers
Projects in Absurd can build and maintain deep memory. As I make the film, the canvas holds an increasingly detailed understanding of the narrative and the characters, what they have lived through, how they feel about it, and how I feel about it.
When I open a new conversation on a chapter’s page, that memory is already there. I don’t have to explain the whole film again. I can begin from my current understanding of the story, and continue building.
The second half
The film was originally going to end with the death of the first emperor. I had already edited most of that story when I realized the ending wasn’t satisfying enough. I wanted to know what happened next. I wanted to explore more of the world and follow the taster further into his life. There was more to be discovered.
The second half hadn’t existed, not even as an idea, for much of the time I was making the first. Now I began to see what could follow: the emperor’s son growing up to become worse than his father, the taster witnessing all of it, outliving both of them and returning home.
The beginning came later, too. When I started, the film opened with the taster already in the palace, his first day on the job. The scenes showing where he came from were created much later.
There is no clean division between pre-production, production and post-production in this strange new medium. I was still writing the film up until the very end.
First-person piece from Devon’s production canvas and its working conversations. Images are working material from The Poison Taster.
Q&A with Devon Johns
Edited from a conversation for length and clarity
- Coming from traditional filmmaking, what drew you to AI?
Devon JohnsAI offers possibilities that are not available in traditional filmmaking. I’m drawn to the unlimited potential it gives me as a storyteller. I can try to capture and execute almost anything I imagine. The scope and scale are much greater, and there are so many more kinds of stories I can tell than in traditional production, where resources can be limiting.
- How has AI changed your filmmaking process?
The main change is that there is no clear division between pre-production, production, and post-production. I can continue writing right up until the end. I might write, create, and edit part of the film, then realize the story needs something I had not originally imagined. I can go back, write it, generate it, edit it in, and discover the next thing that is missing.
Traditionally, you write and lock a script, shoot it, wrap production, and then edit. Unless you can afford expensive reshoots, your options to continue developing the story are limited. With AI those stages become a much more fluid process.
- As a cinematographer, how has AI affected the way you approach images?
I’m not a particularly technical cinematographer; I don’t focus on the details of cameras and lenses. I’ve always followed my instincts. The same goes for AI: either I like an image or I don’t. You can have conversations with the AI about what you like and dislike and hopefully get it onto the same wavelength as you so that it produces more of what you’re looking for.
- The Poison Taster features vast ancient Roman settings and costumes. What would it have required through traditional production?
Traditionally, it probably would have taken close to $200 million to make the movie. And even with that budget, after building your sets you would be locked into what you created. For every set in the film, I probably explored 30 or 40 different options. It is difficult even to comprehend the expense of constructing all the alternative sets, costumes, and scenes just to test them the way I could in Absurd before arriving at the final version.
