An anthology series that follows three women in different decades all living in the same house, as they deal with infidelity and betrayals in their marriages.
Future Associate delivered 130+ VFX shots across the production — approximately a third of which were period driving composites set in a CG 1949 Los Angeles environment, all shot at night in the rain. The balance covered environment work, atmosphere and a range of compositing across the series.
Why Women Kill Season 2 is set entirely in 1949 Los Angeles — and as anyone who has spent time in LA knows, it is a city built around driving. The production needed an efficient pipeline for a significant volume of period driving scenes, all of which presented an obvious problem: you cannot simply go and shoot background plates in 1940s Los Angeles. The streets, the cars, the signage, the storefronts — all of it would need to be created.
The production's solution was to engage a specialist company to build a CG period environment of Los Angeles in Unreal Engine — multiple city blocks covering a range of street types and a more detailed Chinatown environment that served as a hero location. The original intention was to play this content back in real time on an LED volume behind the actors in a stationary car. When that proved impractical on the day, the production pivoted to green screen, shooting the actors inside the car against green and handing us both the Unreal environment and the green screen plates to composite together in post.
Our first task was matching camera angles. Every cut in a driving scene represents a different camera position — driver side, passenger side, two-shot facing down the front of the vehicle, shooting from outside — and for each angle we needed to determine exactly what the background environment should look like and how it should be framed. Crucially, we also matched the film back and focal length of each camera angle, so the depth of field and perspective of the CG background would feel consistent with the lens that captured the actors. Once the angles were locked, we set the travel speed at around 40 kilometres an hour and began animating the world around the car: background vehicles moving through the streets, pedestrians on the sidewalks giving the city life and scale.
For each angle, we rendered out approximately a thousand frames from Unreal as EXRs — over-length plates that went straight into compositing. The time shifting of those plates was handled in comp: each background became a stock element that compositors could slip and reframe in time to match the story beats of the scene. A dialogue-heavy driving sequence cuts frequently — driver, passenger, two-shot, passenger again — and the same background angle might appear multiple times across a scene. The priority when selecting which section of the plate to use in each cut was always keeping the background unobtrusive. These are soft, out-of-focus elements of a 1940s city passing by; the audience's attention needs to stay on the dialogue and the actors inside the car. Anything that catches the eye — an unexpected vehicle entering frame, a sudden change in the streetscape — undermines the scene. The background should feel continuous and believable without ever demanding to be noticed.
All of the driving scenes are set at night in the rain. We added wetness, haze, and wet window elements throughout — water beads blown across the glass as the car moves through the streets. One of the standout shots faces towards Chinatown, red lanterns strung overhead across the full width of the road. Shot at night with a shallow depth of field, those lanterns bloom and disperse their colour through the wet beaded windows — a beautiful image that gives the world genuine scale and atmosphere.
One shot required a tracked camera — a character walking along the road in the rain. We tracked the camera in Nuke and exported the track into Unreal, which was new territory for us. It worked beautifully: Unreal rendered the environment with the tracked camera, that render was handed to compositing as a plate, and grading, atmosphere and wetness were composited on top alongside the foreground. An unusual pipeline for us — our default tools are Houdini and Redshift — but Unreal gave us the efficiency this job demanded, and the client was able to review and approve assets in the real-time engine before they came to us, which kept the process moving.
The limitation of the Unreal pipeline was AOV output — you don't get the range of render passes you'd expect from a modern offline renderer. Our compositors approached those renders the way they'd approach a plate from production: working with what was there, rather than the full suite of passes they might normally have access to.
The rain on set was practical — the actors were genuinely wet, shot in a rain machine. But pulling a clean green screen key means removing everything in front of the green, including the rain falling across it. We keyed out the screens, which meant stripping the practical rain in the process, and then rebuilt it entirely in post using library elements. The rain on the actors was already there in the plate; our job was to restore the rain in the environment around them so the composite felt continuous.
In the wider shots, rain splashes on the roof of the car were added using library elements — small details that sit at the edge of consciousness but do significant integration work, grounding the car in the wet environment and helping the green screen edges read as part of the same world as the background.