Back in 2017, Eran Dinur walked us through the VFX work on The Lost City of Z. Since then, he’s led the visual effects on a wide range of films including Hostiles, Ad Astra, The Trial of the Chicago 7, and The Woman in the Window.
How did you get involved on this film?
I worked with Josh Safdie on Uncut Gems, and he wanted to continue working with me on this movie too.
How did you organize the work with your VFX Producer?
We had a VFX producer (KJ Pilar), who was here in NY and was involved in all day-to-day work, and an executive VFX producer (Allen Maris) in LA, who oversaw the budget and communicated with the studio. The support of both producers helped me focus on the creative and technical aspects and work closely with Josh.
How did you choose the various vendors and split the work amongst them?
We had meetings with quite a few companies. Before joining this movie, I worked as director of training at FuseFX, which was part of the Pitch Black group that includes Folks and El Ranchito. I collaborated with both companies and personally knew many of their key crew, which was a factor in our decision to select them. Another reason we chose Folks and El Ranchito was that they share the same pipeline, which indeed proved helpful later in post. We really liked Folks’s crowd portfolio, and I knew their main crowd expert in Toronto, so we decided to award the crowd to them and the other types of shots to El Ranchito.
The title sequence was kind of its own thing, and we knew that it would need a unique creative process, so we decided to engage a New York company, The Artery, to create it. I have known their owner for years and felt that they have the creative knack for this kind of work.
As the Production VFX Supervisor on Marty Supreme, how did you approach the philosophy of invisible visual effects, especially on a film so grounded in performance and period detail?
To me, it was important to always have as much practical elements as possible in every VFX shot, and to avoid big green screen scenarios. For example, we shot the interior Morosco theater in New Jersey. Logistically it made sense to shoot the exterior scene at the same location. But when we scouted it, we realized that nothing in the environment looks like the NY theater district. I remember telling to Josh and the producers, “this is going to be a giant green screen box”. Exactly what I wanted to avoid. We ended up shooting that scene in the actual Broadway theater area. Sure, we still had to clean up a lot of non-period elements, and we had to add a couple of CG trucks to hide the crowd of onlookers that gathered at the corner. But the street, the buildings, the layout were the real deal. We even got a theater façade in the background that looked so great, we only had to replace the name of the play on the marquee with a period one.
Many of the ping-pong rallies were filmed without a ball—how early was that decision made, and what were the key advantages and risks of removing such a critical physical element on set?
We knew from the start that we’d need to add a CG ball to some of the matches, and we were prepared for that. But the decision on whether to play with a real ball or use VFX was usually done on the day. Table tennis is such a fast and tight game, even top players can’t keep 100% accuracy on multiple takes. Playing without a ball allowed the actors to put all their focus and energy into the performance and allowed Josh to move quicker through the shots and takes. The main risk was timing – even a couple frames too late or too early can make the ball feel too slow or too fast. Thankfully, both Timmy and Koto were amazingly accurate in their timing.
Can you explain how the table tennis coach, playback system, and frame-counting methodology helped you validate the actors’ timing before any VFX work began?
Initially, Diego (the table tennis coach), KJ (the VFX producer) and I planned to use a playback system that would allow Diego to quickly count the frames between each hit and determine if the take was good or needed another run. But the reality on the set made it impossible to use that system – things had to move fast, as we had so much to cover on each day. I was also constantly busy working with the AD’s on positioning the extras for each shot to make sure we optimize the layout for CG crowds. In the end we relied on Diego’s intuition, and it worked well.
When animating the ping-pong ball, why did you choose a fully keyframe-based approach rather than simulation, and how precise did the animation need to be to sell realism?
Table tennis is a game of spin – top spin pulls the ball down and allows the player to send a fast ball that would still land on the table and bounce off with energy. Under spin, which is a defensive hit, slows the ball down and makes it somewhat floaty. Each point was choreographed precisely, and we knew where the ball should go and what kind of spin it carries. Trying to get all this correctly using physical simulations would be completely counterproductive. Not to mention the fact that we often had to “cheat” physics a little to achieve the desired effect. There was a story to tell, and we had to treat the ball as an animated character.
How did you ensure that the ball’s spin, speed, and angle remained physically believable while still matching the actors’ choreography exactly?
It certainly wasn’t easy and often took many iterations to get right. The animators had to get into the “mindset” of a ping pong ball – it’s a tiny, super light ball that is heavily affected by air resistance and spin. As soon as I joined the movie, I started taking lessons with a table tennis coach to get a better “feel” for the ball and the dynamics of the game. I wanted to be well prepared for guiding the animators through.
With tracked virtual cameras, lighting, shadows, and reflections, what were the biggest challenges in integrating the CG ball seamlessly into handheld and multi-camera footage?
We had enough references with a real ball in these environments, and of course we took spherical HDRI’s and measurements for each lighting setup, so matching shadows and reflections was pretty straight forward. The bigger challenge was actually motion blur. It was often tricky to find the right settings, especially when the camera was also moving frantically.
Crowd work appears to be one of the most demanding aspects of the film—what made the Wembley World Championship sequences particularly complex from a VFX standpoint?
We had about 200-250 extras for the world championship, and we had to fill roughly 5000 seats at the Meadowlands Arena (the location we used as Wembley). Some of the games were shot with up to 5 cameras, so it was pretty much a 360 coverage, lots of handheld moves. The lights were focused on the tables and sharply fell off over the bleachers, but the challenge in that environment was integrating the CG within the thick, smokey atmosphere.
For the Tokyo finals, the crowd shows a huge range of ages, costumes, and behaviors—how did you approach character variety and cultural specificity at such scale?
The Tokyo crowd was definitely the most challenging. There was no darkness or atmosphere to hide behind, every detail was visible. We scanned about 30 extras in Tokyo, but even that was not enough for variety, and Folks had to create additional wardrobe variations. Cloth simulation was also a challenge, especially those loose kimonos. The crowd in Tokyo is very active and is really part of the story, so we had to go far beyond a limited library of generic reactions.
How did constant changes in daylight, atmosphere, and crowd energy affect the simulation and rendering of these large-scale audience shots?
The lighting was another big challenge. Every shot had areas of extremely bright direct sunlight right next to dark shadowed areas. And these would be different in every shot. When we shot those scenes, I tried to place few extras in the empty areas in each shot as reference. But getting the integration right was a tricky balance of CG lighting and comp work.
The film includes full 1950s environment replacements such as Times Square, Haneda Airport, and even the Egyptian Pyramids—how did you balance historical accuracy with subtlety so these shots never feel like VFX showcases?
As I mentioned before, the approach was to always have prominent practical elements that would set the tone. For example, in Haneda it was the airplane, the people and the vehicles on the ground, in Egypt it was the camels, the extras and the sand (we shot it on a beach in Connecticut). We shot the night scene at the car dealership in an empty parking lot, with only a few period cars and some string lights. We then created the entire background of neon signs and semi-industrial structures and even added some cars passing by. But having those elements in the foreground made a big difference.
None of these shots read as “establishing shots”. They sell the environment in a subtle way that doesn’t feel like a showcase. For example, you never see the full pyramid, just a piece of it, and Times Square is hinted in the back through lights and marquees. There is an understandable temptation to use VFX to create “wow” moments, and I think Josh wisely avoided that.
Beyond environments and crowds, the film features CG bees, a CG dog, a gas station mayhem sequence, and extensive period cleanup—how did you prioritize and integrate these disparate invisible effects into a cohesive visual language?
My approach was always the same – to try and integrate the VFX in a way that does not call attention. The dog falling off the car is CG, but in the next shot, when we see it running away as the light pole collapses and explodes – that’s a real dog that we rotoed out of a different shot and comped in. The bees are CG but the beehive is a real prop – so it helps the blending. On top of that, small imperfections or seemingly random elements help make the VFX feel more realistic. For example, in one of the bee shots you can see one bee sort of aimlessly crawling on his arm, a little touch that I think helps sell the randomness of reality. It’s the same approach as the production design in this movie – never too clean or perfect, always feeling a bit messy and random.
How long have you worked on this show?
A full year from preproduction to post.
What’s the VFX shots count?
Around 500 shots.
What is your next project?
I am currently doing on-set supervision on another A24 movie. I am also working on two new courses for fxphd and another online school. And I just finished the 2nd edition of The Complete Guide to Photorealism, which will be released in the spring.
A big thanks for your time.
// VFX BREAKDOWNS
// TRAILERS
WANT TO KNOW MORE?
El Ranchito: Dedicated page about Marty Supreme on El Ranchito website.
© Vincent Frei – The Art of VFX – 2026
