With The Conjuring: Last Rites crossing the $500M mark to become the highest-grossing horror film of 2025, Digital Domain stood at the heart of the film’s supernatural terror. Under the supervision of Alex Millet, the studio delivered 425 shots ranging from invisible fixes to hero sequences: Annabelle’s monstrous transformation into a 12-foot stalker, Abigail Arnold’s unnerving smile, the haunted mirror’s shattering hallway nightmare, and an entire Pennsylvania mill town built in CG. Armed with Houdini, Nuke, Maya, Solaris, Photoshop, and V-Ray, the team balanced photorealism with dread, crafting visuals that were frighteningly real without ever slipping into camp. We sat down with Millet to discuss scope, artistry, and the technical pipelines behind conjuring digital horror.

Meet Alex Millet (IMDB | Linkedin), a familiar face in the world of visual effects and someone whose résumé reads like a tour through modern cinema. Currently Visual Effects Supervisor at Digital Domain, Alex began his career as a lighting and look-development artist, later moving through roles as CG Supervisor, Sequence Supervisor and DFX Supervisor.

His filmography spans some of the most recognisable titles in contemporary filmmaking, including The Dark Knight Rises, Inception, Doctor Strange in the Multiverse of Madness and The Electric State. Each project sharpened his eye for realism and storytelling through light, texture and composition, preparing him for his latest haunting venture with The Conjuring: Last Rites (2025).

DP: Alex, can you start by outlining the scope of Digital Domain’s work on The Conjuring: Last Rites?

Alex Millet: The major effects were the Smurl street environment, the hallway destruction and the Annabelle doll sequence. Combined, these represented a little over half of the scope. On top of these, we also had a very complex ceiling replacement in the mirror room, a CG dog, some fire enhancement and along with many other invisible effects such as glasses reflections, TV insert, background replacement, etc. The team grew up to about 150 artists, and production lasted about 8 months. 

DP: Which sequences do you personally consider the centerpiece of your team’s work, and why?

Alex Millet: The Smurl environment was a big part of our scope. We recreated the entire street surrounding the Smurls’ house and the looming factory. The environment was seen in multiple sequences at different times of day, each with different lighting and “life” in the neighbourhood. The hallway destruction was another big challenge because of all the destruction that we had to create. And, of course, Annabelle being so iconic, the sequence where she comes to life was an important one.

DP: Horror relies on timing and subtlety. How did you collaborate with the director and editorial team to ensure the VFX didn’t overwhelm, but instead amplified the tension?

Alex Millet:Scott Edelstein was the VFX Supervisor on the production side and our main link with the director, Michael Chaves. Working with Scott and Michael was amazing; they were open to ideas and understood what was possible, which made it feel very collaborative.

The goal for us is always for our effect to disappear and blend with the on-set photography to complement the look and feeling the director created. We match every detail as closely as we can; down to the subtle amount of wind in the atmosphere. The production team did a great job anticipating our needs while shooting, which made it a lot easier to integrate our image and keep our effect non-distracting.

DP: The Annabelle transformation is one of the film’s signature moments. What were the creative conversations around taking an iconic prop and making it terrifying at a larger-than-life scale?

Alex Millet:We started by building a CG version of the original doll, then turned it into the 12-foot version. As the giant version came to life in the hallway, the iterative process began. The 12-foot version of Annabelle went through a few rounds of redesign to become the scary character that the director wanted to see. Once the new look was found, we received a concept and started rebuilding our asset and worked on a new post viz. Annabelle needed to look determined and threatening as she came out of the room, yet not totally in control of her body.

As the animation team took over, some aspects they focused on were Annabelle’s silhouette, matching some of the poses established during the concept phase, and the transformation. The transformation needed to feel abrupt, painful, and scary. We scaled different limbs at different times across the cut while working very closely with our CFX team to make sure the jerky motion did not break the dress simulation or make it distracting. The lighting also played a key role in the scare, keeping things dark yet revealing just enough to feel Annabelle’s presence, until the final jump scare with the flashlight.

DP: Abigail Arnold’s smile is disturbingly subtle. How much back-and-forth was involved in refining something so specific, yet so essential to the character’s menace?

Alex Millet: We were lucky that the Director, Michael Chaves, is quite VFX-savvy, and provided us with a concept of the look and smile he wanted Abigail to have. The goal was to match his concept and bring her to life. We started with recreating the concept for different lighting scenarios, then developed a range of smile variations, ranging from neutral to wide extended smiles. The director was then able to select the right performance and the range of the smile that worked for the shots.

A lot of the subtlety is in everything around the smile. The mouth only moves so much, but the way the cheeks, nostrils, eyelids or even eyebrows, compress and deform, all contribute to the realism and expression. Once we had selected range we refined the motion further, and the compositing team integrated the new face into the shot. We used this process for different shots, all in different lighting environments.

DP: Horror effects can easily cross the line into unintentional comedy, the dreaded “absurdist 80s VHS fail.” How do you dial a CG effect into the sweet spot where it reads as horrifying, not parody?

Alex Millet: Here again, blending into the tone and vision established by the practical photography is always the most important for the effects. Our goal is always to make sure the VFX are as seamless as possible, and not distracting.

Other than matching the look, lighting and ambience of what was done on set, that also means making sure that any animation does not go over the top, and breaks the tension. The Abigail smile was a great example, we could get a much wider range for a smile but instead, the more subtle ones, and in some cases barely moving at all, kept her looking disturbing.

DP: Any tips for your peers on how to land that “maximum horror” effect faster in iteration, instead of burning weeks on trial and error?

Alex Millet: Hiding things is often a lot scarier than trying to explicitly create something scary. As VFX artists, we’re very analytical of our images and want to see every detail to make sure it’s perfect, but in this case, we needed to hide part of the image and let things fall into darkness. We’re still just as concerned with every detail; it’s more of an extra step: now that we know how it looks, what does it look like way darker? What parts are gone, and which ones are left. We, then, let the audience fill in the blanks.

DP: You’ve mentioned the Pennsylvania mill town environment. Can you walk us through how the team extended sets into a fully believable small town, while keeping continuity across day and night shots?

Alex Millet: We knew we were going to see the street in many different shots, angles and lighting scenarios; we just didn’t know what these were going to be until the edit stabilized. So it started with concept work, then committed pretty early on to building the entire street as a fully renderable CG asset.

We started with a rough 3D layout of the neighbourhood, then focused on the street the Smurls’ house was on. We built the houses in a modular way, like Lego blocks, and split them between the different floors and roofs, so we could easily swap the different pieces and create a lot of variations. This allowed us to start working on our assets to a high level of detail early on, and also prevented having to redo work if, for example, Scott or Michael wanted to move a particular house, or change the roof on another.

Once the VFX sup & director were happy with the look in some early development shots, we worked on the different lighting environments and added life with props, cars, atmosphere, etc. So having the street fully built made it easy to maintain continuity across the shots, or break it if we wanted to do so for creative reasons. Around the CG street, we used 2.5D, mixing matte painting with supporting geometry for the factory and background, and lastly, the far background and sky were 2D matte painting.

DP: Beyond the big hero moments, you delivered a large volume of invisible effects, reflections, vehicles, ceiling replacements. What kind of pipeline ensures consistency across such a wide range of tasks?

Alex Millet: Having a scalable pipeline is key. Many individual effects were only for a few shots, while others, like the Smurl street or the hallway destruction, spanned over 100+ shots. Having a single but scalable pipeline allows us to be nimble and iterate quickly to get started, but then scale up to handle any volume of shots. If things change, it also prevents us from having to go back and build things differently or with a different approach.

DP: Let’s talk specifically about the mirror sequence. How did you design your glass-break pipeline for that scene?

Alex Millet: We had to create the shattering glass in the hallway sequence as well as the crack that’s on the mirror throughout the movie. Glass is really unforgiving in the sense that it needs to be fully built for all the reflections and refractions to behave correctly; there is no shortcut. Our asset team built a CG version of the mirror first, ensuring the glass layer had the correct thickness and reflection behaviour.

Then for the shattering, the FX team fractured the glass with an evolving crack. Fracture simulations don’t always result in the cleanest geometry and with glass, any imperfection can result in unwanted highlights. So after the glass was fractured in FX, we went back to the asset team and spent more time refining the geometry to make sure everything behaved correctly and looked as real as possible. For the static cracks, we had a different approach and started with a 2D concept of the overall shape, then sculpted it directly into the geometry.

DP: The haunted mirror and hallway destruction sequence demanded complex simulation. What tools and workflows did you lean on to ensure the scene felt both physically credible and cinematically heightened?

Alex Millet: The hallway sequence was a challenge. We started with building a CG version of the hallway that had to be water-tight to work with the FX simulations. Our assets are mainly built in Maya for modelling, Mari for texturing, then Houdini & VRay for Lookdev.

Once our asset matched the set, we added some extra geometry behind the walls and floor, to be revealed during the destruction and added material variations to the debris. The next step was to animate the mirror moving through the wall and start working on the look of the broken wood and floorboard simulation.

Our FX team, working in Houdini, did different iterations of breaking wood and drywall to find that rhythm and dynamic pop in the destruction. We also played with different amounts of smoke, either slowly filling the room or shooting from the impacts. We then went back to lookdev to create some procedural materials for the additional geometry.

That’s all the inside faces that didn’t exist in the original model and were created once the geometry was fractured. Some of the finer effects, like smoke or fine particles, went directly from the FX team to the compositing team, but the main fractured geometry was passed to the lighting team to render with the mirror and the rest of the hallway. The lighting team rendered everything with VRay and Solaris in Houdini, then passed it to the compositing team to put all the elements together in Nuke.

DP: How many iterations did it take to nail the timing and behavior of the breakage, and how did you balance simulation randomness with director’s notes?

Alex Millet: The biggest challenge was to tie in the propagating crack, which was an FX simulation, with the main cracks. The main cracks, under the actor’s hands, had a very specific art direction to look like they originated from the hand. They also needed the right timing to happen when the actors touched the mirror. We laid out the main impact points first, then matched the speed of the simulated propagating crack to tie everything in.

Once the timing worked, we tweaked the geometry of our propagating crack once more so the lines would connect with the main impacts. The final effect was meant to look like a single event, moving across the entire glass but with specific impact points tied to the actor’s performance..

DP: Looking back, what do you consider the biggest technical achievement of your team on The Conjuring: Last Rites?

Alex Millet: The hallway sequence was an interesting technical challenge; it came in late in the schedule, but required a full rebuild of the hallway and a heavy amount of FX destruction in almost every shot. The mirror was plowing through the walls and floors, and we needed to maintain some consistency in the destruction across the sequence.

We also had to be very careful of the actors’ reflections, as we sometimes had to replace the mirror with a CG one for the animation and destruction to work, but we needed to maintain the actors’ reflections. Overall, the biggest challenge of the project was the aggressive timeline and wide scope of work; we had a big environment build with the Smurl street, a lot of FX simulation, some animation and character work with Annabelle and Abigail, a huge oner shot, and the list goes on. The team did an amazing job delivering this project on time and without any outstanding notes.

The post When Annabelle Grew Up: The Digital Domain Horror Pipeline first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

Can a Visual Effects studio survive only on film work? “Absolutely not!”

Scott Ross is an important part of the history of Visual Effects. He was manager at ILM, Senior VP at Lucasfilm, Chairman & CEO of Digital Domain and one of the Founding Members of Visual Effects Society. He summarized his experiences and a part of the industry’s history in his book “Upstart”. In our episode we trace a line from the cloistered halls of ILM and Lucasfilm through the founding of Digital Domain, across the rise of digital pipelines and into today’s inflection point of Agentic AI.

Scott’s path into film wasn’t linear: Music studies, audio engineer, then Bay Area post with One Pass before Lucasfilm recruited him. When he arrived at ILM in the mid-1980s, the reality contradicted the Oscars on the shelf: ILM was losing money and morale was low. The corporate layer above him prized brakes and cost cutting over progress and acceleration.

Scott sketched ILM at a crossroads. The facility’s reputation masked cashflow issues and a flattening culture. His first months were spent earning the trust in a “Hell’s Angels”-tight shop that viewed him as a young “video” guy; an outsider. Once in, he ignored the brake pedal people and ran ILM as if it were his own: Buying new gear, fixing the org chart, restoring transparency and diversifying revenue to smooth the feature-film boom-bust cycle, opening up opportunities for commercial and location-based attractions.

Soon after he found himself on a crossroad to either stand behind his team and support their efforts for a fair compensations and bonuses that they were promised or accepting the games at play. Waging his job for what he believed was right he found himself looking for another job. The problem was that he was the general manager of one of the biggest visual effects studios on the planet, no one else wanted to replace their GM while Scott’s reputation wasn’t of someone who played the cooperate games.

“A fixed contract to do something that’s never been done before, giving the facility all of the responsibility and none of the authority, is a model of madness.”

Scott Ross

Three-legged stool

Over the years Scott defined the concept of the “three-legged stool” as the key ingredients of a successful business. Remove even one leg and the shop tips and the business fails. The three legs consists of:

• Technology: Technology is the key to staying relevant in quality and production speed. ILM’s analog-to-digital pivot foreshadowed today’s shift toward AI-assisted craft.
• Finance: A diversified revenue is key for any business but especially for visual effects as there are only a handful of production houses. From commercials, attractions, episodic productions to visualisations those had different field while often have better margins.
• Creativity: Keep the core creative decision-makers close to the tools, but align authority with responsibility. Scott’s ILM anecdotes reveal how often errors are organisational rather than technical.

“Follow your bliss, but be brutally honest about your lane, and plan your rent. You can’t survive without profit. If it doesn’t make money, it’s not a business.”

Takeaways

Scott’s stool test is still the fastest sanity check for any show or facility. If technology dazzles but cash burns, the shop tips. When finance is tight but creative authority is outsourced, the frames suffer. If creativity is strong but tech lags deliverable specs, you ship compromise. The “fixed bid for the impossible” contract structure is as dangerous now as it was then. As AI moves from assistive to agentic, we need to write scopes, schedules, and authority trees that assume fewer hands on more decisions and still protect image integrity and human sanity.

Finally Scott warned against believing one’s own hype: if you want to be a core storyteller, develop those muscles; if not, your skills will be valuable across medicine, manuals, advertising, and beyond, especially as visual communication saturates every industry.

Do you want to read this for yourself? The Book “Upstart: The Digital Film Revolution. Managing the Unmanageable” by Scott Ross and Joanne O`Brien-Levin.

The post “A Model of Madness” Scott Ross Why VFX Fails as a Business first appeared on DIGITAL PRODUCTION and was written by Alexander Richter.

For The Fantastic Four: First Steps, Digital Domain contributed nearly 400 shots spanning character builds, creature effects, and dialogue-driven sequences. At the centre of the work stood The Thing, Baby Franklin, and the retro-futuristic robot Herbie, with Digital Domain tasked to balance comic-book authenticity and emotional believability.

DP: Could you walk us through Digital Domain’s overall contribution to The Fantastic Four: First Steps. Nearly 400 shots across multiple characters and sequences?

Phil Cramer: We had the unique opportunity bringing some of the most Marvel iconic characters to life. Our main focus was creating the Thing, Herbie, and Baby Franklin. We also helped work on Johnny Storm, Mr Fantastic’s experiments, and Sue’s powers. A lot of our work was centered around the Baxter Building, our heroes’ headquarters. Our focus was often the character and dialogue driven story beats, leaning on Digital Domain’s character experience.

DP: The Thing was an asset shared across multiple vendors. How did you establish consistency in performance and look across studios, and what role did your Masquerade3 pipeline play in this?

Phil Cramer: Over the years, sharing between vendors has more and more become a standard, especially on Marvel shows. Fantastic Four, however, was a special example of camaraderie unlike anything I have seen before. From day one, each vendor’s VFX supervisor would join a weekly design meeting with Ryan Meinderding, Head of Visual Development at Marvel Studios. After The Thing’s look was locked down, I was tasked with finding the best capture and FACS combination to suit every vendor’s needs.

Once our Motion Capture Supervisor Connor Murphy identified Masquerade as our main facial solver for the show, Digital Domain processed and provided over 1500 post-viz assembly shots to The Third Floor, ILM, Sony Picture Imageworks, Framestore, and our internal team. This became a critical building block for our main character, The Thing, and allowed everyone to start from the same point.

In comparison, for Thanos, we only shared the asset itself, but the facial work would fall on each vendor’s proprietary system together with the huge development burden to come up with the same look twice.

DP: Masquerade3 is mentioned as a key innovation. Could you explain the major advancements of this version compared to previous pipelines?

Phil Cramer: Masquerade3 has significantly advanced our facial capture pipeline. The biggest change is the markerless approach, which completely streamlines the entire process. This means the actor can suit up, put on a helmet camera and jump straight into the scene. Normally, we would need to apply the facial markers at least 1 hour before the shoot, with touch ups as needed throughout the day. During Covid, masks would constantly smudge the markers. Due to this, we would constantly need to rely on re-calibrating the system based on the subtle changes to the marker layout. On She Hulk, for example, we were drawing in daily calibrations.

In addition, we are now able to run solves on the entire footage captured and run it via a command line blindly. In the past, we would receive “selects” per shot and have facial artists process this data. Now, we are literally running our system on the ENTIRE dataset overnight. This mass processing allowed the FACS system to be used by the Third Floor for postviz, as well as all the final vendors.

A nice little side development for The Fantastic Four: First Steps occurred during our additional photography shoot. Ebon Moss-Barach had to have a beard due to his other commitments. With a markered system, this makes the data useless (in part because the beard covers the markers or makes the markerset inconsistent). With Masquerade3, this proved to be no issue and we were able to solve at the same quality all Ebon shots with a beard.

DP: Processing 60 hours of facial footage sounds like an enormous data challenge. How was the workflow structured between capture, processing, and integration into final animation?

Phil Cramer: Over the past few years, our facial processing has evolved drastically. While we used to process the data on a per shot basis, we now are batch processing all captured data per character. This means, we no longer have facial artists spend significant time on each individual shot. We focus on an overall correlation between the actor and the character via animation. Once this is established, we can consistently produce results at very high quality, but solved blindly via the command line.

We would mimic every night the entire facial delivery from Marvel with matching animation files. This was critical to allow all other finals vendors as well as the postviz vendor, The Third Floor, to utilize the Masquerade3 facial solve provided by our team. Once this became a critical step for each vendor, we set up an assembly team under Connor Murphy. He processed and set up over 1500 “layout” files with facial and body mocap. This helped keep all data consistent and allowed for the most advanced starting point.

DP: How did the team balance the comic-book heritage with the cinematic portrayal of a believable infant?

Phil Cramer: Franklin had to be a believable baby at the core of it all. He was meant to be as realistic as a baby as possible, with no special powers. This was so important that we conducted a full test with Matt Shackman about two years before the release of the film. We even used my own newborn son, Cosmo Cramer, as the test baby to better understand how to consistently and convincingly replace a real infant..

We knew consistency of the babies was the highest priority. We captured endless performances of the baby actor portraying Franklin. We captured a wide range of baby motions in 4D for training and as potential face replacements. In the end, we used all sorts of techniques to create Franklin, but the dominant approach focused on head replacements with a full CG head.

DP: For Baby Franklin, you used both Charlatan and a baby-friendly scanning booth. Can you explain how those workflows were adapted to make the CG infant believable while working within (ethical and) practical limits of filming with real babies?

Phil Cramer: From the very beginning, the baby was high on every list of show priorities. We needed an effective solution to capture the best possible performance, while also respecting the limited on-set time for baby actors. Because of this, we had around 30 stand-in babies that VFX needed to make look like one single, consistent baby character.

Capturing a precise performance from a baby is practically impossible as they cannot be directed. And, we did not want the audience distracted by an artificial baby’s performance. Director Matt Shackman came up with the notion of accidental acting. Every day, we’d set up the same critical scenes with the hero baby and stunt actors in the hope to get the right beat on an emotional level. This was a very important aspect for the resurrection sequence after the final battle, when Sue stops breathing, and the Fantastic Four along with Baby Franklin gather around her, believing they’ve lost her.

For most of the other scenes, we needed a versatile baby setup with a strong CG asset at its core paired with lots of training and reference material. Clear Angle helped us set up a baby rig that allowed a young baby to be sitting in it for hours at a time, with easy access for the parents. We needed to ensure we can capture the highest level of detail under the lowest possible light to protect the baby’s eye.

DP: You also worked on fire, invisibility, and ultrasound effects. How did you approach combining procedural effects with narrative-specific ones like the pregnancy ultrasound?

Phil Cramer: Sue’s power from the get go had some really interesting, classical looking FX work. The highlight for us was the baby ultrasound. Our FX team tried to generate pleasing looking, but highly detailed internals. We wanted to reveal the baby in many layers, but it could not be too medical. So, we tried a more ethereal approach to show the inside of the human body, focused on refractions and different lens aberrations. The beauty of this scene was critical. In fact, it was one of the first development tests we did on the show utilizing a stunt woman with a prosthetic belly.

DP: HERBIE seems to combine hard-surface modeling with comedic character animation. What kind of rigging and animation workflow did you build for him, and how did you integrate “mechanical logic” with slapstick humor?

Phil Cramer: Herbie is such a wonderful part of the Fantastic Four universe and we wanted to make sure he fit right into the 60s style world. On set, there was a matching puppet that was used to show the actor’s where Herbie is at any given moment and that they could interact with. A puppeteer was also present at all times.
Most of the Herbie motions were based on the overall puppeted performance. We would normally track the overall body motion, but then enhance it in animation. This added to the realism, as he kinda was real. I love characters with limitations, and Herbie is one just like that. It helps ground a character into realism. Herbie was quite limited to his ball motion (yes he could fly, but he would barely utilize his skill). Herbie ended up being a lovely house robot that anyone would like to have.

DP: Nine costumes for The Thing, plus facial hair simulation. How did your team manage asset complexity and cloth/hair simulations across so many variants of basically a pet rock with anger issues?

Phil Cramer: The Thing’s character is not like many others in comic book lore. He is not happy with his super powers, in fact he wants to hide them. He is also ashamed of his new look. He is constantly trying different ways to hide his body and face, even to the point of growing a beard. I love this kind of a character. It shows the imperfections we so often miss in superheroes.

Our Asset Supervisor Ron Miller made this process feel easy. Once we established a good collision mesh for the Thing’s body, we dressed all the costumes against it. Production helped us a great deal by having a stand in actor in a prosthetic suit available for any shot. He would always wear the correct costume tailored to The Thing. This really allows you to ensure you match the properties of the design, rather than make a CG costume.

DP: The film leans heavily into a 1960s retro-futuristic design language. How did that aesthetic influence your VFX choices, from Baxter Building interiors to character look development?

Phil Cramer: All the FX were meant to have the same 60s feel as the actual world did. So we set out to look for traditional looking FX from that era, but done with today’s tools. Sue’s refraction edge break up falls in this category. We wanted to find an effect that would look like classical color blocking, while being executed in a modern way.

For the baby proofing sequence in the Baxter Building, we had to look for all sorts of period correct baby proof devices. Lots of fun for the team to find classical fire detectors, etc.

DP: With characters like The Thing and Baby Franklin, there’s always the “uncanny valley” problem. What were the biggest creative decisions to make them feel both comic-book authentic and emotionally relatable on screen?

Phil Cramer: The Thing himself was grounded in the artwork of Jack Kirby. The goal was a character reflecting the design and spirit of the original Thing by Jack Kirby. We wanted to find ways for the lighting to shape him similar to the 2D drawing, i.e. very bold light and shadow play.

Franklin was its own story. He is not purple or made out of rock. Everyone has seen a baby before, so everyone will be able to judge. Our goal was to approach every shot in a unique way, and to not give away our hand. As far as performance was concerned, we always needed to ensure the baby behaved like a baby in that age… so no superhero things. Franklin could only do what a 4 – 6 month old baby could do… which is close to nothing :)

DP: Were there specific sequences where VFX had to “step back” to let the actors carry the scene, and others where VFX had to “take over” entirely?

Phil Cramer: I feel Digital Domain’s scope of work is dominantly in the “step back” category. Meaning, we ended up having a lot of the story driven sequences of all our main characters. It wasn’t about anyone in particular, but the Fantastic Four team learning and planning together as a family. The Thing behaved pretty much like an actor. He was not about action, but acting. So our job on set was to stay out of the way. We wanted to encourage the creative process between these four brilliant actors, and make them forget about the technology.

DP: Marvel films often balance spectacle with continuity across the MCU. How did you ensure your work fit seamlessly into the larger Marvel visual language while still creating something fresh for Fantastic Four?

Phil Cramer: The Fantastic Four: First Steps was intentionally set in a different multiverse to feel free and fresh. We tried to make each effect and character look and feel different from the regular MCU. Having worked on so many MCU features and episodicals, this was a nice, new experience. It all started with the 60s look of Matt Shackman’s world. From there on, every aspect was meant to break the typical approach. As an example, the Things had to be based on Jack Kyrbi’s original work. He was meant to bring the original drawings to life… in a modern form.

DP: The Thing’s beard became both a character trait and a technical challenge. Was this a case where creative direction and technical innovation directly influenced each other?

Phil Cramer: Our Thing went from growing a stubble to a full fledged beard. Funny enough, so did the actor. We initially asked him to shave to accommodate the technology, but later allowed a full beard in support of his role on Season 3 of The Bear.

Growing the beard gave The Thing a lot of character, but, in turn, made the emotions harder to sell for the animation team. Our Animation Supervisor Frankie Stellato and his team did an amazing job still selling the same range of emotions, even with his face hiding under a full beard. Interestingly enough, a beard of that size is quite static. We ended up using a rigid body sim for the beard and it became part of our regular CFX dailies.

DP: You collaborated closely with other VFX vendors. What worked best in terms of pipeline sharing, and what lessons can the industry learn from this cross-studio cooperation?

Phil Cramer: Sharing is caring, especially on a Marvel show. Fantastic Four provided a very open environment for all companies to get together and share ideas. Early on, we all collaborated closely with Ryan Meinderding in a very open design process with all VFX supes involved. Marvel then tasked me with creating a common capture list for all the vendors, to ensure everyone gets what they needed to make their characters work.

This led us to evaluating Masquerade3 as the main solver for all vendors. It ensured everyone started with the same animation and the same base FACS shapes in their rigs. We did group Thing animation reviews with Production VFX Supervisor Scott Stockdyk, which really helped dial the character in. I think we all learned a lot from that process and I have a lot of respect for my fellow supervisors at ILM, Sony and Framestore. It was a unique experience and I hope we maintain this open approach on future projects.

DP: Looking ahead, do you see more projects adopting this kind of multi-vendor, shared-pipeline approach, especially around proprietary tools like Masquerade3?

Phil Cramer: I hope so. I feel this process helped everyone involved. In general, we need to avoid over-capturing/processing. These large scale VFX productions can be improved by being smarter about the capture and processing. Rather than having each vendor replicate the same ingest and training costs, this streamlines the process, and without really reducing the creative. It is like digital housekeeping.

DP: What would be your recommendations for artists or studios attempting similar workflows, shared assets, markerless capture, large-scale character development?

Phil Cramer: The assembly phase, led by Connor Murphy, was the crux of this shared production. It served as a central hub through which all motion traveled through, like a universal layout pass with both body and facial mocap.

How can you replicate that as a student or younger artist? I feel with the dawn of AI, lots of fun tools have become accessible. There are different forms of AI capture to simplify the body performance. Unreal offers some of the finest body and face retargeting. Try to replicate the core steps to capture a character.

DP: And finally: what’s next for you and your team? Are there upcoming projects you can tease for us?

Phil Cramer: My team is going from small to big… After focusing for two years on babies, we are now working with Godzilla and his friends :)

The post Digital Domain on Fantastic Four: From Baby Mocap to Bearded Rock first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

Blackmagic Design has introduced a 16 TB Media Module of its proprietary media module for the URSA Cine 12K LF, doubling the capacity of the previous 8 TB module. The new 16TB Media Module installs into the URSA Cine’s media bay and uses M.2 SSDs over four PCIe lanes, enabling professional productions to record 12‑bit Blackmagic RAW and H.264 proxy files at speeds up to 1 432 MB/s—enough for sustained 12K at 24 fps. The module requires in‑camera formatting and ships with a rugged plastic case and standard one‑year warranty.

Bulk storage, cloud‑style workflows

The 16 TB module retains the features of its 8 TB sibling: it works with Blackmagic’s Media Dock for hot‑swappable file transfers, and supports high‑speed 10G Ethernet connectivity, allowing live file access over network or to Blackmagic Cloud Mini setups integrated in the camera. These features position it as “a Cloud Store Mini built into your camera”.

Pre‑order pricing

Available now for pre‑order, the 16 TB module is priced at US $2 965, versus the original 8 TB unit. Delivery timing may vary by region; check with Blackmagic’s reseller network for local availability.

Filmmakers shooting high‑resolution, extended takes can double recording capacity without swapping modules, reducing downtime. Those using dual CFexpress may still prefer existing formats, but for extended 12K workflows this is a significant upgrade.

Final thoughts

This 16 TB offering strengthens Blackmagic’s proprietary media ecosystem for the URSA Cine 12K LF, marrying high‑speed M.2 SSD performance with on‑set networked workflows. As usual, new gear should be tested before deploying in production environments.

The post BMD ups the ante with 16 TB Media Module for URSA Cine 12K LF first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

Thunderbolts*, directed by Jake Schreier, arrived in theaters May 1st, 2025, following a lineup of Marvel operatives turned uneasy team. It blends grounded espionage thrills with comic‑book spectacle—think cloak‑and‑dagger meets superhero showdown.

DP: Ryan, Thunderbolts* doesn’t exactly “whisper” visually. When Marvel called, what was the scope of work that landed on your desk—and what made you say “yes”?

Ryan Wilk: Digital Domain has collaborated with VFX Supervisor Jake Morrison on a few projects, so when he asked us to join the show, it was a no brainer for us. The scope of sequences this time around were a bit less flashy or grand than our usual fare (at least by Marvel standards). We were asked to help on two primary sequences, the vault fight and elevator shaft escape. Both sequences are quite grounded in their action (think Jason Borne rather than Captain America), but there was still quite a lot of VFX lifting to be done, it just may not be obvious to the viewer initially.

DP: Marvel films have a reputation for being VFX-heavy, but also iterative. How early was Digital Domain involved—and how much changed between first concept and final comp?  

Ryan Wilk: We started our asset builds in the spring of 2024, so pretty early on. We then got news that the studio wanted to show a good chunk of our sequence at Comic-Con, which landed in July 2024.  This really turned up the heat, but the silver lining is we got to finalize quite a few of our assets and successfully establish the visual vocabulary of the Ghost phasing effect. Please don’t ever tell marketing this, but the Comic-Con push ended up really helping us establish things for the final film.

In regards to things changing between concept and final comp, usually, that is quite the journey on Marvel films. But this time things were fairly stable on our sequences. Of course, there were changes and creative exploration, but even if you go back to the original previs and stunt vis, they are quite similar in spirit to the final edits. Our previs team actually collaborated closely with director Jake Schreier, providing a clear visual blueprint early in the process, with a focus on fast and iterative development that supported his creative vision. This included mapping out the continuous progress of an early version of the elevator sequence.

DP: This vault isn’t your average storage space—more like a brutalist trap disguised as a filing cabinet. When you first saw the previs, what was your reaction to the creative and technical ambition of this sequence? 

Ryan Wilk: Early in the bidding phase, when you see a sequence like this you always ask yourself “this is cool, but how much can they actually shoot”?  Typically on Marvel films, our first impulse is to build everything from floor to ceiling. Be prepared for all CG shots. Make sure those walls are ready for destruction in case Hulk comes crashing in for a cameo (joking, that was never mentioned).

But on this film from the beginning, creatively (and budgetarily), the filmmakers had a plan to shoot these grounded action sequences, introduce the primary characters of the film and use VFX to really punch up the intensity and dynamics of the fight. And that held mostly true. We built out the vault floor fully in CG, along with some hero props and boxes. But we didn’t do a floor to ceiling build. For the extensions needed (hallways and higher ceilings), we used a DMP with a Nuke projection setup.

DP: Give us a sense of scale—how many shots did Digital Domain handle for Thunderbolts*, how big was your team, and how long were you on the project?

Ryan Wilk: We ended up producing 160 shots in the final film, and around 200 if you factor in marketing versions. We started on the show in Spring 2024, so almost an entire year. Team size varied across the schedule, but about 200 people touched the show in one capacity or another.

DP: The vault sequence manages to stay physical, while hiding hundreds of digital tricks. How did you approach the VFX so that it never shouted “look, we’re CG”?

Ryan Wilk: To stay grounded, our mantra became less is more. From spark hits to explosions, we would often start big and find ourselves dialing things back before approval. In cases where a heavier CG hand was needed, i.e., full digital doubles, we iterated heavily in animation to make sure the physics and motion of the characters didn’t jump out as implausible. And our hard surface & digi-double assets were top notch, so our CG renders really integrated seamlessly into the final product (with some help from comp, of course).

And finally, we could always go back to the plates. Even when replacing a character, we had a live-action reference to dial towards. The filmmakers stood by the photography they captured, so we were not pushing things around unnecessarily.

DP: Did real-time tools like Unreal Engine or LED wall setups play any part in your VFX planning, or was this all traditional plate-based post?

Ryan Wilk: Traditional plate-based post for our sequences.

DP: That vault set was 100 feet long. You extended it digitally with vents, walls, and connecting hallways. Can you walk us through how you scanned, rebuilt, and extended the space?

Ryan Wilk: Production used LiDAR to capture the set, which we then cleaned up and used for our needs. The hallways, vents and upper walls were extended using a Nuke projection setup, while the floor and key props were built out fully in CG. For the elevator environment, we also started with LiDAR and texture photography from the live-action set, but ended up tweaking the lookdev a bit to get the final look.

DP: Paper flying, sparks flaring, boxes tumbling—what was simulated and what was practical in the vault chaos?

Ryan Wilk: The primary vault fire flood (when the vents expel fire downward) was fully simulated in Houdini. We also simulated boxes, paper, and small debris to help with the realism. Ron Howard’s “Backdraft” was a great film for reference and inspiration for our explosions and fire. 

We also did the explosive arrow simulations that hit Walker’s shield.  In both cases (shield and flood), it was a matter of iterating to find the right balance of scale, density and timing to remain believable.

DP: In the Yelena throw shot, you combined Florence Pugh’s face with a stunt performer’s full-body take. Can you break down how Charlatan evolved for this? And how hard was it to keep hair and costume alignment believable?

Ryan Wilk: This was probably our trickiest shot. To get a seamless handoff, it really took all available tools we had. We started with an articulate track of the stitched Yelena performance and rendered her digital double with hair and cloth sims. Using our proprietary tool, Charlatan, we completed a face replacement utilizing other plates and takes of Florence as source material. 

And lastly, our team had the idea to feed the Charlatan tool a hero frame and target frame for Yelena’s body itself, and the tool attempted to generate content to fill the gap. All of the elements (plate, CG render, and Charlatan content) were combined and finessed heavily in comp. And in addition to all this, the environment also had to be fully reconstructed to work across both takes seamlessly and have Yelena interact with the CG set. We hope that when people watch this shot, they don’t even realize it is a VFX shot!

DP: Let’s talk Ghost. Her phasing effect must’ve gone through eleventy iterations. What was the final approach, and how did you balance the ‘techy’ visual logic with audience readability?

Ryan Wilk: In this iteration of Ghost, she has more control of her powers than in previous films, so her phasing was meant to be more intentional. For her motion, we tried to utilize her plate performance as much as possible, but played with the timing and spacing to create the ‘leaves’ of Ghost. We created the ‘leaves’ in animation, and then rendered them with varying shutter angles to provide comp with options from streaking effect with long shutters to sharp non-motion blur renders .

Our comp team then did the final sweetening, using chromatic aberration and varying levels of defocus and opacity to create the final look. It was quite a creative challenge, because we were not just adding the Ghost effect to each shot arbitrarily, we were experimenting and using the Ghost effect to help enhance the action of the stunts. Each shot with Ghost phasing was uniquely crafted from Anim to Lighting to Comp to get the necessary look for buy off.

DP: When the incinerator kicks in, things explode fast. What went into building the fire, pressure, and structural damage—without turning it into a Michael Bay moment?

Ryan Wilk: I think it really comes down to iterations. For the flood in particular, it did take a few rounds to find the right speed and volume to maintain believability. It was imperative to use past movies with live action explosions as reference to help determine the scale of the simulations. Once we were good with the fundamental simulation of the fire flood then we ran the secondary additional sims to drive elements like smoke, embers,  boxes, papers, small debris to further help with the realism.

DP: Were there any custom in-house tools or pipeline tweaks developed specifically for Thunderbolts*—whether it was for sim setups, render optimization, or automation?

Ryan Wilk: Nothing in particular, other than maybe stretching the abilities of our face-swapping software, Charlatan, into new areas. A happy accident to discover that Charlatan can be used more like a glorified morphing tool to stitch plate characters with digi-doubles and vice versa.

DP: One thing about this vault brawl—it’s messy in a good way. Did you help shape the stunt choreography visually, or were you purely there to augment the physical fight?

Ryan Wilk: Most of the choreography is faithful to the original stuntvis, but we do like to think we contributed to it in small ways throughout the sequence. In one shot, for example, there was a bit of a pause in the stunt moves while Ghost and Taskmaster were fighting. The filmmakers wanted Ghost to be phasing, but because Taskmaster wasn’t actively attacking at that moment, the phasing felt unjustified. 

So we replaced Taskmaster’s arm with a CG version, so that her sword swipes for Ghosts’s head. This allowed us to use the phasing as a dodge, and another ‘close call’ moment. There were little moments like this throughout, and the filmmakers were always open to our pitches to help improve and tighten the action of the fights.

DP: You retimed and repositioned Taskmaster’s arm mid-punch to sell weight. That’s oddly poetic. How much of the fight was rhythm-adjusted in post to maintain impact?

Ryan Wilk: We did things like this throughout the sequence, some very minor, some requiring a bit more surgery.  And this is not to take away anything from the stunt work, it was very good. But adding just a little extra snappiness with the help of VFX makes everything feel just that much more intense.

DP: The camera move during Yelena’s throw is a standout—no fast cuts, no motion blur cheats. Was that shot the hardest to comp—or the most fun?

Ryan Wilk: Both!

DP: There’s a subtle photorealism to your digi-doubles in – they don’t scream “CG.” How do you approach the uncanny valley now? More shaders, better rigs, better lighting—or all of the above? 

Ryan Wilk: All of the above. We are always trying to improve our digis, but for years now, if you were to take a look at our digital double t-pose turntables, you would agree they are highly photoreal. It is always fun when you show a side-by-side to someone and they can’t tell which is the reference and which is CG. But I think the challenge in sequences like this really comes down to the animation and motion. The human eye is just so keen when it sees things move unnaturally. So, superhero movies have always had this particular challenge…you have humans doing non-human things, and making that appear real to the human eye/brain is always tricky.

In Thunderbolts*, while there were a few moments where we were defying physics a bit, most of it was very grounded. The viewer could buy that something like this could be done by a very athletic human. And I think that contributed to us being able to get things to really sit in nicely.

DP: When you’re designing FX like Ghost’s phasing, do you think in emotional beats—like “this one should feel desperate” or “this flicker should feel like anger”?

Ryan Wilk: I don’t know if it got that nuanced, but it certainly was fun coming up with ways for her phasing to help the action. Let’s add a narrow miss here, Let’s have her pop up behind someone at just the right moment there for the ultimate sucker-punch, etc.

DP: What’s your current favorite “small-but-crucial” tool in your pipeline—something you think every VFX team should use, but maybe overlook?

Ryan Wilk: Well you were lucky enough to get the VFX producer for this interview instead of a supervisor, so am I allowed to answer with an excel formula? I’ll go with “Index match”.

DP: What’s the one thing you’d tell a VFX team starting their first Marvel-scale show?

Ryan Wilk: Be prepared for changes, show your work often, and be honest if you are in the weeds.

DP: Everyone’s talking about neural rendering lately. Did Gaussian Splatting, NeRFs or similar volumetric techniques sneak into Thunderbolts*—or are we still in “experimental, not production” territory?

Ryan Wilk: We’re actively investigating the Gaussian Splatting that Chaos introduced with VRay 7 for some relighting techniques and better lighting all around within our workflow. But for Thunderbolts* we stuck with the tried and true HDRs, using projections on combinations of LiDAR and layout geometry for our lighting.

DP: The vault is one of the most grounded VFX sequences Marvel’s done in years. What’s the shot in there that quietly makes you smile every time?

Ryan Wilk: There will always be one shot that makes me smile, and that is a seemingly simple shot where the vault doors all slam closed as the camera whip pans around the room. The doors in this shot are CG, and despite being one of the first shots turned over, it took ages for us to get this right. We had wedges of differing door drop speeds (simulated and key framed), different mechanisms for how the door panels move against each other. Numerous iterations on lookdev to get the door texture and lighting just right. Sometimes the most seemingly easy visual effects task can become a struggle. But we all get a good laugh out of it now.

DP: What’s next for you and Digital Domain—if it’s not NDA’d into a black hole?

Ryan Wilk: Not allowed to say quite yet, but Digital Domain has quite a few exciting projects going for the rest of 2025.

The post Thunderbolts* Vaults: Digital Domain on Marvel’s Most Grounded Brawl first appeared on DIGITAL PRODUCTION and was written by Bela Beier.