My AXE Commercial Died in a Budget Meeting. Five Years Later, I Brought It Back with AI
In 2021, director Cassandra Brooksbank pitched a large-scale AXE commercial through The Martin Agency - a subway-set disaster comedy in which a man's enormous orange puffer jacket inflates from body heat until it drenches an entire train car. The concept was shelved on budget grounds. Five years later, Brooksbank used it as a test case for Absurd, a collaborative AI filmmaking platform built around a persistent project memory that tracks characters, locations, and creative decisions across a production.
The process Brooksbank followed was deliberately close to her live-action workflow. She started with a written director's treatment, which she then used Absurd's image tools to develop visually - generating character portraits, location options, and storyboard frames instead of wrestling with Photoshop composites. She cast a full ensemble of around 20 distinct synthetic characters, locking references for each one before generating any footage. Different models handled different tasks: GPT Image for casting portraits and environments, Uni for wardrobe and multi-angle character reference sheets, and Ideogram for editing clothing onto locked character references. Absurd's shared memory carried those decisions forward so she was not re-explaining the cast and setting with every new prompt.
For video generation, Brooksbank used Seedance 2.5, working in multi-shot sequences of three to five shots at a time, reviewing 480p drafts before rendering approved takes at 1080p. She treated AI direction the way she treats actor direction - giving characters specific motivations rather than asking for generic emotional states, and establishing physical rules for how effects should behave. The edit was cut in Premiere, with finishing handled by cinematographer Andrew Fletcher in DaVinci Resolve. Fletcher graded individual shots in a wider color space to preserve flexibility, then used Filmbox Pro to add 250D film grain and halation to soften the footage's digital sharpness.
Brooksbank is candid about what AI did not change: the fundamental decisions about framing, timing, performance, and story still required human judgment. What the tools did change was the cost and time required to execute those decisions - the entire storyboard took under a day, the video generation ran about two days, and the project reached a finished cut without a physical shoot. Her account is notable less for the technology involved than for the discipline she brought to it, treating every stage of an AI production as she would a real one and finding that the results tracked closely to the film she had originally planned on paper five years earlier.
