Table of Contents Show
LOOKSfilm brought Overmind Studios in as the sole VFX vendor for all six episodes of A World Divided, a historical drama following six real figures from World War II through the 1960s. The series blended archival material with recreated drama and moved across locations from New Mexico and China to Russia, Israel and French Algeria.
Shot in Germany, Belgium, Luxembourg, and Poland, the project demanded constant geographic reshaping and period-authentic cleanup to sell multiple continents and decades, all while working to rolling episodic deadlines.
Tobias Kummer was VFX lead on A World Divided at Overmind Studios. He supervised on set on key shoot days, built the project pipeline and delivered the work alongside André Gerhardt as a two-person VFX team. (Site | IMDB | Linkedin | Xing)
DP: How did you come onto A World Divided, and what was the overall scope?
Tobias Kummer: LOOKSfilm brought us in as the sole VFX vendor for all six episodes. We’ve worked with them before, so there was already trust there. The series was a historical drama following six real figures from World War II through the 1960s. It wove archival material together with recreated drama and jumped across locations from New Mexico and China to Russia, Israel and French Algeria.
André Gerhardt and I handled the VFX work as a two-person team, with me also covering on-set supervision on key days. Across the season, we delivered around 125 shots in total.
DP: What were the main categories of VFX work across the season?
Tobias Kummer: Most of the work fell into a handful of categories. The biggest chunk was geographic transformations and set extension work. The show was shot in Germany, Belgium, Luxembourg, and Poland, but it needed to represent locations across multiple continents, so we were constantly reshaping the world. We made the Pustynia Błędowska desert in Poland read as New Mexico in some scenes and as Israel in others. We replaced window views with believable locations and extended sets to fit both the period and the geography.
Period cleanup was the other constant. Removing anachronisms took up a steady portion of the workload, such as power lines, phone masts, wind turbines, modern signage, contemporary vehicles, basically anything that didn’t exist between the 1940s and 1960s. Some removals were straightforward paint-outs, while others required proper tracking and reconstruction for moving shots, especially given the anamorphic lens characteristics to account for.
Beyond that, there was a lot of environmental and atmospheric work. We added snow to summer shoots where winter was required, enhanced practical smoke and explosions with additional elements, and built the nuclear explosion sequence. There were also the smaller, problem-solving shots that always appeared on a show like this, including window inserts, moon replacements, damage work, retouching visible wig seams and removing on-set facilities from reflections. Many shots weren’t just one technique. A single exterior might have needed cleanup, set extension and atmos all layered in one comp.
DP: What constraints shaped the work most?
Tobias Kummer: What shaped the work was the schedule and the fact that we were a two-person team delivering an episodic shot count with rolling deadlines. Episodes moved into colour on a rolling basis, which meant we had to work sequentially and finish each episode properly before moving on to the next. We couldn’t cherry-pick shots across the whole season when we needed breathing room. There was a clear rhythm to the work. However, we had to maintain strict discipline.
“What shaped the work was the schedule, and the reality of being a two-person team delivering an episodic shot count with rolling deadlines.”
DP: How did the anamorphic photography affect VFX, especially cleanup?
Tobias Kummer: The anamorphic photography added complexity in places you might not expect. Distortion and breathing made tracking more challenging than it would have been with spherical lenses, particularly on handheld or more dynamic shots. We had everything from locked-off tripod plates to dolly moves, cranes, and handheld coverage, so even simple cleanup could have become complicated if you wanted it to sit cleanly while preserving the anamorphic feel.
DP: What did on-set supervision change for you in practical terms?
Tobias Kummer: On-set supervision helped in practical ways. Sometimes it was about making sure we had clean plates and tracking markers. At other times, it was flagging modern details before they became post problems. Or it was a small camera placement decision that saved hours of rotoscoping later. We had an initial planning phase with the DOP and directors in May 2023, and on-set supervision ran through the shoot from May to July.
Then there was a long gap. We didn’t receive VFX plates until May 2024, nearly a year after wrapping, because production was working through the edit in the meantime. That delay reinforced something I always try to stress. Write everything down during on-set supervision. No matter how certain you are that you’ll remember something, you won’t a year later. Good notes made it possible to pick the work up again without having to guess.
DP: How did the schedule work once plates started arriving?
Tobias Kummer: Once plates started arriving, we had roughly two weeks per episode, regardless of complexity. VFX production ran for over 12 weeks across the six episodes. Shot complexity varied widely. Some cleanups were quick, while sequences with multiple iterations took longer, but the template system helped. The first shot of a new type might have taken time to establish the approach, but similar shots afterwards moved much faster.
Approvals were efficient. Most shots were approved in versions 1 or 2, and only a handful needed additional revisions. When revisions did happen, they were usually about creative direction rather than technical fixes, which came back to communication and understanding intent early.
DP: How did you split responsibilities as a two-person team?
Tobias Kummer: André and I divided the work in a clear way, but we kept shot ownership flexible. I handled on-set supervision, project planning, client communication and pipeline setup, and I also did shot work as well. André focused on shot execution and handled the motion graphics. We didn’t rigidly assign shot types. It was more fluid. Whoever was free took the next thing that needed doing. Because we were working remotely, communication was essential. We stayed connected throughout the day, and if one of us got stuck, a quick call and a screen share often solved what might have taken hours otherwise in minutes.
DP: What did you use for project management?
Tobias Kummer: For project management, we used Ramses by RxLaboratory. That was our central truth for what was in progress, what was out for review, and what was approved.
I wanted to avoid the usual folder wrangling that came with episodic work, so I built a small Fusion plugin that connected directly to the Ramses API. In practice, it meant that when you opened a shot, the setup was already done. Correct plates, correct naming, correct version number, correct saver paths, and comp settings that matched the show spec. It wasn’t glamorous work, but it prevented mistakes and kept us moving. Templates were the other foundation. I kept them rigid where mistakes were expensive and flexible where a shot needed freedom.
DP: Can you break down the plate specs and colour pipeline?
Tobias Kummer: Plates came from Avid as 4K MXF containers in DNxHR HQX 12-bit with a 1.8 anamorphic squeeze, in ARRI Wide Gamut 4 and LogC4.
We composited in scene-referred linear, so the input linearisation was standardised. Output was standardised too. For review, we rendered previews in Rec.709 using the client reference LUT, so the work looked close to the intended look. Finals went out in ARRI Wide Gamut 4 and LogC4 so the grade could treat VFX shots like any other plate. The LUTs were helpful for context, but we couldn’t bake them into the finals without limiting the colourist’s downstream options.
DP: Why Fusion Studio, and what tools mattered most?
Tobias Kummer: All compositing happened in Fusion Studio, because it was what we knew and because the node-based approach was genuinely flexible under episodic revision pressure. When notes came back, you could restructure a comp non-destructively. Insert processing, reroute parts of the tree, try alternatives, without rebuilding from scratch.
Fusion’s tracker and planar tracker held up well on the anamorphic plates, which mattered for both cleanup and set extension work on moving shots. Magic Mask helped as well. It wasn’t perfect in every case, but it saved us more than once from manual roto. We supplemented with a handful of Reactor tools and some custom tools we’d built, but the aim was to rely on Fusion’s core toolset and maintain a consistent pipeline for broadcast delivery.
We also used Blender for 3D elements and set extension geometry, EmberGen for pyro and atmospheric simulations, SynthEyes for more complex camera solves and Affinity Photo for detailed paint work. All rendering was done locally rather than on a farm, which gave us more control and faster iteration.
DP: Pick one shot that best represented the season. What was it, and how did you approach it?
Tobias Kummer: If I had to pick one shot that summed up the work, it was a vast desert establishing shot where we turned the Polish Pustynia Błędowska into an Israeli desert landscape. The plate was shot on a crane mount, so there was a slight shake, and the camera racked focus between foreground and background. In the original plate, you could see the forest at the edges of the location, which didn’t sell the intended geography at all.
This was a case where on-set supervision paid off. Had we gone with the initial framing, we would have had the actors’ heads and the car crossing the horizon line, which would have meant rotoscoping them out to replace the background. I asked for a slight adjustment to the camera placement so everything stayed below the horizon. That saved hours of roto later and made the background replacement much cleaner.
In Fusion Studio, we used a few point tracks to stabilise and reapply the move, and we paid close attention to the rack focus because anamorphic lenses can shift and breathe, causing the background to drift if you ignore it. We built a matte painting from Israeli desert reference photos for the mountainous backdrop. We added blowing sand using an EmberGen simulation we had created and reused across multiple desert shots, and we added heat haze with a simple Fast Noise feeding a Displace node. The focus rack was matched by hand with keyframes. The shot was approved in version two.
DP: What were the key lessons you took from delivering this as a small team?
Tobias Kummer: If I had to distil it down to what kept us on time while maintaining quality, it came down to a few habits. First, learn some scripting, Python or whatever your tools support, and automate the repetitive, error-prone parts of the job. The time savings compounded quickly, especially for a small team, and you don’t need to build for massive scale to benefit.
Second, organise everything and document it. When you’re juggling episodes, you can’t waste time on version mix-ups or folder hunting, and the year gap between shoot and plate delivery made that even more important. Lastly, don’t waste time being stuck. Fresh eyes solved problems faster and a quick screen share could save hours. And honestly, sometimes the best decision was to stop and sleep. If you weren’t making progress, you’d often solve it faster in the morning.
