A week in the life of an Unreal environment build — from camera plan to tech check — and why the last two days matter more than the first three.
When we quote a green screen shoot, there’s a line in the schedule that clients tend to skim past: allow 5–7 working days for environment build. It sits there looking like admin. It isn’t. That week is where the shoot is actually decided — whether you’ll be able to move the camera the way you want, whether the key will hold on the day, whether we spend the afternoon shooting or the afternoon troubleshooting.
We’ve been asked more than once what we’re actually doing during that week. It’s a fair question, and the honest answer is that it looks nothing like building a render. Here’s how the week runs at our stage in North Wales, and why the tail end of it is the part that saves you money.
Day One: The Brief Is a Camera Plan, Not a Mood Board
Almost every environment brief arrives as pictures. A Pinterest board, a couple of location stills, a reference frame from something the director loved. That’s a useful start, but it tells us nothing we need in order to build.
What we actually need is a camera plan. Not a shot list in the storyboard sense — a physical description of what the camera will do. Is it locked off, or on a slider? Does it move through the space or just look at it? Where is the talent standing, and how far back does the world need to be convincing? Does anyone ever turn around?
The reason this matters is that a virtual set is not a place, it’s a volume of believability. If the camera never moves and the talent stays on a mark, we can build something extraordinarily rich in a very narrow wedge and spend nothing on the rest. If the camera is going to drift laterally across a two-metre slider, that wedge widens, parallax becomes real, and everything behind the subject has to hold up to being seen from more than one angle. Same brief, twice the build.
So day one is a conversation, not modelling. We take the reference images and translate them into a plan: how many distinct setups, what the camera does in each, how tight the framing goes, whether the background needs to react to the camera at all. We settle the frame rate and delivery spec here too — it’s a five-minute conversation on day one and a genuinely irritating problem on day six.
By the end of the day we know what we’re building and, more usefully, what we’re not building.
Days Two and Three: Grey First, Pretty Later
We block the whole environment in grey before anyone picks a texture.
This is the bit that surprises people. You’d think a virtual set starts with the look, but starting with the look is how you end up with a beautiful environment that runs at eleven frames a second. Real-time is a budget, and the budget is spent in the order you make decisions.
So days two and three are geometry and scale. Blank grey boxes, correct dimensions, correct sight lines. We put a virtual camera in the space with the same lens we’ll be shooting on and walk the moves from the day-one plan. It’s remarkable how often something obvious falls out at this stage — the doorway the camera would clip, the window that puts a hard highlight exactly where the talent’s shoulder will be, the “big open warehouse” that turns out to feel cramped at a 32mm equivalent.
Fixing that in grey costs an hour. Fixing it once the set is dressed costs a day.
Once the blocking survives the camera moves, we start dressing — and we dress outward from the frame. The wall behind the talent gets real attention. The corridor visible through the doorway gets less. The city thirty metres beyond the window is, more often than not, a card. We’re not being lazy; we’re spending the frame budget where the camera is actually looking, because everything we save here we get back later as headroom.
We build in two or three revision rounds across this stretch. Clients see the greybox, then the first dressed pass. Feedback on a greybox is cheap and specific (“the ceiling’s too low”) in a way that feedback on a finished render never is.
Day Four: Building for the Keyer, Not for the Render
Here’s where a green screen environment build diverges hard from anything you’d make for post.
Our final image is composited live. The talent is keyed in real time in Assimilate Live FX and the composite goes to the monitors as we shoot — that’s the whole point of the room. Which means the environment isn’t a background plate that a compositor will lovingly integrate next week. It’s one half of a marriage that has to work instantly, in front of the client, with no undo.
Practically, that changes what we build.
We build the environment’s lighting as a lighting plan, not just a look. Every key source in the virtual world gets a real-world counterpart on the stage — direction, colour temperature, hardness and relative intensity all noted, because our LED reference panels need to reproduce that wrap on the talent for the composite to sit. A gorgeous virtual sunset that we can’t match with practical light is a virtual sunset that will make your presenter look pasted on. We’ve written before about matching talent to a virtual world on green screen, and this is where that job actually starts — five days before anyone stands on the mark.
We also watch the edges. High-contrast, high-frequency detail directly behind where a head will be is a keyer’s nightmare — fine hair against busy foliage is how you get a crunchy matte in front of a client. If the plan puts the talent against a detailed area, we either soften it with depth of field in the environment or move the detail. Nobody in the room ever notices. That’s the point.
And we lock the colour. The environment gets built and viewed through the same pipeline it’ll be shot through, so what we approve on Thursday is what appears on the monitor on Monday. We’ve gone into that pipeline in detail before — the short version is that an environment signed off in the wrong colour space is an environment you’ll be arguing about on shoot day.
Day Five: The Tech Check Nobody Books and Everybody Needs
The last day isn’t creative at all, and it’s the one we’d cut last.
We put the whole chain together and run it: the Blackmagic Pyxis 6K, the Mo-Sys StarTracker Max on its 15mm rod rail, Live FX, the environment, the reference lighting, the Rec 709 monitoring. Then we do the boring things.
We check the tracker’s calibration against the actual lens we’ll be using, and we walk the camera through the real moves from the day-one plan while watching the composite — not the environment on its own, the composite. We confirm that everything on the floor is locked to the same clock, because a virtual set that drifts by a frame is a virtual set that swims, and swimming is the single most reliable way to make a good composite look cheap. If genlock and frame sync are new territory, we wrote a plain-English guide to it.
And we deliberately try to break it. We push the camera further than the plan requires. We go wider than the widest framing. We ask what happens if the director wants to swing round to the other side of the desk, because on the day someone always does — and the difference between “yes, give us ten minutes” and “no, sorry” is whether we found the edge of the world on Friday or on Tuesday.
By the time you walk in, the environment has already been shot. Just not with you in it.
What’s Next
We’re currently working on compressing the front half of this week rather than the back half. Photogrammetry and Gaussian splat capture are getting good enough that a location we can physically get to might become a usable backdrop in a day or two rather than three — we set out where we think splats actually earn their place a couple of weeks back, and this is exactly the arithmetic we meant. Days two and three get shorter. Days four and five don’t, and shouldn’t.
We’re also building out a library of pre-lit base environments — a handful of rooms and exteriors already colour-locked and already carrying a documented lighting plan — so that a project which fits one of them can skip most of the build entirely and go straight to a tech check. Not every job needs bespoke. A lot of jobs need fast.
If you’ve got a shoot in mind and you’re trying to work out whether it’s a five-day build or a five-hour one, come and see the room. Book a studio visit and we’ll walk you through a live composite, break it, and rebuild it — it makes a lot more sense standing in it than it does written down.
This Fortnight in VP
Brompton’s Hydra recalibration doubles SP Studios’ daily shooting capacity — Red Rock OPS recalibrated 315 Absen tiles from different production batches in three days on site, and the studio went from favouring the flat section of its wall to using all of it. A good reminder that most VP capacity problems are calibration problems wearing a costume.
Mo-Sys publishes a buyer’s guide to choosing a virtual production system — Their argument is that platform architecture matters more than any individual feature, and it’s hard to disagree. Worth reading even if you’ve already bought, because it’s a decent checklist for what to ask when you’re hiring a stage rather than building one.
Blackmagic Camera 10.2.2 brings phase-detect autofocus and 10-minute pre-record to the Cinema Camera 6K — This closes the gap to the PYXIS 6K, which got the same PDAF work earlier in the summer. We run a Pyxis, and reliable autofocus on a tracked camera is a bigger deal for VP than it sounds — it’s one less thing pulling focus away from watching the composite.
Mo-Sys completes its recapitalisation and enters a new phase — Following a restructuring process in July, the business and assets moved into a new company structure with the CEO personally backing the acquisition. StarTracker is core kit for a lot of small studios including ours, so continuity here matters to more people than the headline suggests.
IBC2026 runs 11–14 September at the RAI Amsterdam — Virtual production sits across Halls 6, 7 and 14, and Netflix’s Anna Mallett is speaking on how VP, cloud post and AI-assisted workflows are reshaping production for streaming. If you’re going and fancy comparing notes, get in touch.
