By Randi Altman
Good versus evil. It’s a story that’s been told time and time again, but Amazon’s Good Omens turns that trope on its head a bit. With Armageddon approaching, two unlikely heroes and centuries-long frenemies— an angel (Michael Sheen) and demon (David Tennant) — team up to try to fight off the end of the world. Think buddy movie, but with the fate of the world at stake.
In addition to Tennant and Sheen, the Good Omens cast is enviable — featuring Jon Hamm, Michael McKean, Benedict Cumberbatch and Nick Offerman, just to name a few. The series is based on the 1990 book by Terry Pratchett and Neil Gaiman.
As you can imagine, this six-part end-of-days story features a variety of visual effects, from creatures to environments to particle effects and fire. London’s Milk was called on to provide 650 visual effects shots, and its co-founder Jean-Claude Deguara supervised all.
He was also able to talk directly with Gaiman, which he says was a huge help. “Having access to Neil Gaiman as the author of Good Omens was just brilliant, as it meant we were able to ask detailed questions to get a more detailed brief when creating the VFX and receive such insightful creative feedback on our work. There was never a question that couldn’t be answered. You don’t often get that level of detail when you’re developing the VFX.”
Let’s find out more about Deguara’s process and the shots in the show as he walks us through his collaboration and creating some very distinctive characters.
Can you talk about how early you got involved on Good Omens?
We were involved right at the beginning, pre-script. It’s always the best scenario for VFX to be involved at the start, to maximize planning time. We spent time with director Douglas Mackinnon, breaking down all six scripts to plan the VFX methodology — working out and refining how to best use VFX to support the storytelling. In fact, we stuck to most of what we envisioned and we continued to work closely with him throughout the project.
How did getting involved when you did help the process?
With the sheer volume and variety of work — 650 shots, a five-month post production turnaround and a crew of 60 — the planning and development time in preproduction was essential. The incredibly wide range of work spanned multiple creatures, environments and effects work.
Having constant access to Neil as author and showrunner was brilliant as we could ask for clarification and more details from him directly when creating the VFX and receive immediate creative feedback. And it was invaluable to have Douglas working with us to translate Neil’s vision in words onto the screen and plan out what was workable. It also meant I was able to show them concepts the team were developing back in the studio while we were on set in South Africa. It was a very collaborative process.
It was important to have strong crew across all VFX disciplines as they worked together on multiple sequences at the same time. So you’re starting in tracking on one, in effects on another and compositing and finishing everything off on another. It was a big logistical challenge, but certainly the kind that we relish and are well versed in at Milk.
Did you do previs? If so, how did that help and what did you use?
We only used previs to work out how to technically achieve certain shots or to sell an idea to Douglas and Neil. It was generally very simple, using gray scale animation with basic geometry. We used it to do a quick layout of how to rescale the dog to be a bigger hellhound, for example.
You were on set supervising… can you talk about how that helped?
It was a fast-moving production with multiple locations in the UK over about six months, followed by three months in South Africa. It was crucial for the volume and variety of VFX work required on Good Omens that I was across all the planning and execution of filming for our shots.
Being on set allowed me to help solve various problems as we went along. I could also show Neil and Douglas various concepts that were being developed back in the studio, so that we could move forward more quickly with creative development of the key sequences, particularly the challenging ones such as Satan and the Bentley.
What were the crucial things to ensure during the shoot?
Making sure all the preparation was done meticulously for each shot — given the large volume and variety of the environments and sets. I worked very closely with Douglas on the shoot so we could have discussions to problem-solve where needed and find creative solutions.
Can you point to an example?
We had multiple options for shots involving the Bentley, so our advance planning and discussions with Douglas involved pulling out all the car sequences in the series scripts and creating a “mini script” specifically for the Bentley. This enabled us to plan which assets (the real car, the art department’s interior car shell or the CG car) were required and when.
You provided 650 VFX shots. Can you describe the types of effects?
We created everything from creatures (Satan exploding up out of the ground; a kraken; the hellhound; a demon and a snake) to environments (heaven – a penthouse with views of major world landmarks, a busy Soho street); feathered wings for Michael Sheen’s angel Aziraphale and David Tennant’s demon Crowley, and a CG Bentley in which Tennant’s Crowley hurtles around London.
We also had a large effects team working on a whole range of effects over the six episodes — from setting the M25 and the Bentley on fire to a flaming sword to a call center filled with maggots to a sequence in which Crowley (Tennant) travels through the internet at high speed.
Despite the fantasy nature of the subject matter, it was important to Gaiman that the CG elements did not stand out too much. We needed to ensure the worlds and characters were always kept grounded in reality. A good example is how we approached heaven and hell. These key locations are essentially based around an office block. Nothing too fantastical, but they are, as you would expect, completely different and deliberately so.
Hell is the basement, which was shot in a disused abattoir in South Africa, whilst heaven is a full CG environment located in the penthouse with a panoramic view over a cityscape featuring landmarks such as the Eiffel Tower, The Shard and the Pyramids.
You created many CG creatures. Can you talk about the challenges of that and how you accomplished them?
Many of the main VFX features, such as Satan (voiced by Benedict Cumberbatch), appear only once in the six-part series as the story moves swiftly toward the apocalypse. So we had to strike a careful balance between delivering impact yet ensuring they were immediately recognizable and grounded in reality. Given our fast five-month post- turnaround, we had our key teams working concurrently on creatures such as a kraken; the hellhound; a small, portly demon called Usher who meets his demise in a bath of holy water; and the infamous snake in the Garden of Eden.
We have incorporated Ziva VFX into our pipeline, which ensured our rigging and modeling teams maximized the development and build phases in the timeframe. For example, the muscle, fat and skin simulations are all solved on the renderfarm; the animators can publish a scene and then review the creature effects in dailies the next day.
We use our proprietary software CreatureTools for rigging all our creatures. It is a modular rigging package, which allows us to very quickly build animation rigs for previs or blocking and we build our deformation muscle and fat rigs in Ziva VFX. It means the animators can start work quickly and there is a lot of consistency between the rigs.
Can you talk about the kraken?
The kraken pays homage to Ray Harryhausen and his work on Clash of the Titans. Our team worked to create the immense scale of the kraken and take water simulations to the next level. The top half of the kraken body comes up out of the water and we used a complex ocean/water simulation system that was originally developed for our ocean work on the feature film Adrift.
Can you dig in a bit more about Satan?
Near the climax of Good Omens, Aziraphale, Crowley and Adam witness the arrival of Satan. In the early development phase, we were briefed to highlight Satan’s enormous size (about 400 feet) without making him too comical. He needed to have instant impact given that he appears on screen for just this one long sequence and we don’t see him again.
Our first concept was pretty scary, but Neil wanted him simpler and more immediately recognizable. Our concept artist created a horned crown, which along with his large, muscled, red body delivered the look Neil had envisioned.
We built the basic model, and when Cumberbatch was cast, the modeling team introduced some of his facial characteristics into Satan’s FACS-based blend shape set. Video reference of the actor’s voice performance, captured on a camera phone, helped inform the final keyframe animation. The final Satan was a full Ziva VFX build, complete with skeleton, muscles, fat and skin. The team set up the muscle scene and fat scene in a path to an Alembic cache of the skeleton so that they ended up with a blended mesh of Satan with all the muscle detail on it.
We then did another skin pass on the face to add extra wrinkles and loosen things up. A key challenge for our animation team — lead by Joe Tarrant — lay in animating a creature of the immense scale of Satan. They needed to ensure the balance and timing of his movements felt absolutely realistic.
Our effects team — lead by James Reid — layered multiple effects simulations to shatter the airfield tarmac and generate clouds of smoke and dust, optimizing setups so that only those particles visible on camera were simulated. The challenge was maintaining a focus on the enormous size and impact of Satan while still showing the explosion of the concrete, smoke and rubble as he emerges.
Extrapolating from live-action plates shot at an airbase, the VFX team built a CG environment and inserted live action of the performers into otherwise fully digital shots of the gigantic red-skinned devil bursting out of the ground.
And the hellhound?
Beelzebub (Anna Maxwell Martin) sends the antichrist (a boy named Adam) a giant hellhound. By giving the giant beast a scary name, Adam will set Armageddon in motion. In reality, Adam really just wants a loveable pet and transforms the hellhound into a miniature hound called, simply, Dog.
A Great Dane performed as the hellhound, photographed in a forest location while a grip kept pace with a small square of bluescreen. The Milk team tracked the live action and performed a digital head and neck replacement. Sam Lucas modeled the head in Autodesk Maya, matching the real dog’s anatomy before stretching its features into grotesquery. A final round of sculpting followed in Pixologic ZBrush, with artists refining 40-odd blend shapes for facial expression.
Once our rigging team got the first iteration of the blend shapes, they passed the asset off to animation for feedback. They then added an extra level of tweaking around the lips. In the creature effects phase, they used Ziva VFX to add soft body jiggle around the bottom of the lips and jowls.
What about creating the demon Usher?
One of our favorite characters was the small, rotund, quirky demon creature called Usher. He is a fully rigged CG character. Our team took a fully concepted image and adapted it to the performance and physicality of the actor. To get the weight of Usher’s rotund body, the rigging team — lead by Neil Roche — used Ziva VFX to run a soft body simulation on the fatty parts of the creature, which gave him a realistic jiggle. They then added a skin simulation using Ziva’s cloth solver to give an extra layer of wrinkling across Usher’s skin. Finally they used nCloth in Maya to simulate his sash and medals.
Was one more challenging/rewarding than the others?
Satan, because of his huge scale and the integrated effects.
Out of all of the effects, can you talk about your favorite?
The CG Bentley without a doubt! The digital Bentley featured in scenes showing the car tearing around London and the countryside at 90 miles per hour. Ultimately, Crowley drives through hell fire on the M25, it catches fire and burns continuously as he heads toward the site of Armageddon. The production located a real Bentley 3.5 Derby Coupe Thrupp & Maberly 1934, which we photo scanned and modeled in intricate detail. We introduced subtle imperfections to the body panels, ensuring the CG Bentley had the same handcrafted appearance as the real thing and would hold up in full-screen shots, including continuous transitions from the street through a window to the actors in an interior replica car.
In order to get the high speed required, we shot plates on location from multiple cameras, including on a motorbike to achieve the high-speed bursts. Later, production filled the car with smoke and our effects team added CG fire and burning textures to the exterior of our CG car, which intensified as he continued his journey.
You’ve talked about the tight post turnaround? How did you show the client shots for approval?
Given the volume and wide range of work required, we were working on a range of sequences concurrently to maximize the short post window — and align our teams when they were working on similar types of shot.
We had constant access to Neil and Douglas throughout the post period, which was crucial for approvals and feedback as we developed key assets and delivered key sequences. Neil and Douglas would visit Milk regularly for reviews toward delivery of the project.
What tools did you use for the VFX?
Amazon (AWS) for cloud rendering, Ziva for creature rigging, Maya, Nuke, Houdini for effects and Arnold for rendering.
What haven’t I asked that is important to touch on?
Our work on Soho, in which Michael Sheen’s Aziraphale bookshop is situated. Production designer Michael Ralph created a set based on Soho’s Berwick Street, comprising a two-block street exterior constructed up to the top of the first story, with the complete bookshop — inside and out — standing on the corner.
Four 20-x-20-foot mobile greenscreens helped our environment team complete the upper levels of the buildings and extend the road into the far distance. We photo scanned both the set and the original Berwick Street location, combining the reference to build digital assets capturing the district’s unique flavor for scenes during both day and nighttime.
Before and After: Soho
Mackinnon wanted crowds of people moving around constantly, so on shooting days crowds of extras thronged the main section of street and a steady stream of vehicles turned in from a junction part way down. Areas outside this central zone remained empty, enabling us to drop in digital people and traffic without having to do takeovers from live-action performers and cars. Milk had a 1,000-frame cycle of cars and people that it dropped into every scene. We kept the real cars always pulling in round the corner and devised it so there was always a bit of gridlock going on at the back.
And finally, we relished the opportunity to bring to life Neil Gaiman and Douglas Mackinnon’s awesome apocalyptic vision for Good Omens. It’s not often you get to create VFX in a comedy context. For example, the stuff inside the antichrist’s head: whatever he thinks of becomes reality. However, for a 12-year-old child, this means reality is rather offbeat.
Randi Altman is the founder and editor-in-chief of postPerspective. She has been covering production and post production for more than 20 years.