Hollywood in a box: what is lost when every world is built in the studio

What “Gray Box” means here and what virtual production really is

“Gray Box” is explicitly a journalistic metaphor in this piece, not an official industry term. It refers to neutral production halls, closed soundstages, green- and blue-screen studios, LED volumes, and fully controlled virtual production spaces in which the final world is created digitally. Hollywood does not call these spaces that; the metaphor stands for a trend, not for a label.

Three basic terms help with getting started. According to the FilmLA study, a soundstage is a building or part of a building that is usually isolated against outside noise and natural light and meets municipal certification criteria (FilmLA, Soundstage Study). Green screen and blue screen are single-color surfaces for so-called chroma keying: based on color information, a specific color range is selected and isolated, the background is made transparent in post-production, and replaced by a new background through compositing (Adobe, What is chroma keying?). Green is preferred because it does not normally occur in skin or hair tones, so the software can easily separate performers from the background – as long as no one wears green clothing; when green motifs or costumes appear, a blue screen is recommended instead for strong contrast (Adobe).

At the heart of modern virtual production is In-Camera Visual Effects (ICVFX). Here performers stand not in front of a green cloth but inside a curved LED volume that displays a digital world. According to Epic Games' technical documentation, the method relies on a mix of LED lighting, live camera tracking, and real-time rendering with so-called off-axis projection, with the aim of creating the finished image directly in the camera and avoiding compositing against a green screen (Epic Games, In-Camera VFX Overview). The image divides into two zones: the inner frustum is the section the camera sees; it follows the camera movement and produces genuine parallax, so the impression of a real place emerges. The outer frustum, the rest of the wall, remains static and serves as a dynamic light and reflection source for the physical set (Epic Games).

For the perspective to be correct, camera tracking measures the position of the camera in real time, usually via optical infrared markers, feature tracking, or inertial sensors, often in combination (Epic Games). A game engine like Unreal renders the world live; systems like nDisplay synchronize multiple computers, and a genlock prevents image tearing between wall and camera (Epic Games). A Virtual Art Department builds the digital locations previously defined by virtual location scouts and in previsualization; digital set extensions and virtual lighting complement physical sets. Importantly: virtual production is not the same as generative AI. AI is at most a supplementary part of individual workflows, as with Netflix's Eyeline group, which explicitly describes VFX, virtual production, AI-assisted tools, and research as separate, combinable building blocks (Netflix, Eyeline).

The gain: control, continuity, and less risk

The strongest advantage of the volume is control. Cinematographer Greig Fraser, who used the technique on “The Mandalorian” and parts of “The Batman,” stresses the value of the right light at the right time: if you are in the right place at the right time, you hardly need artificial light, and a correctly built volume can – with the right time of day and sun intensity – deliver the same effect (Filmmaker Magazine). This control leads to typical production consequences that vary in strength from project to project: constant light, more plannable shooting days, and lower weather risks. Because filming takes place in the studio, travel may be eliminated, dependence on shooting permits may decrease, and sensitive productions can be more easily shielded from the public – all of these are possible effects, not guaranteed outcomes. Digital environments can also be swapped out without a physical change of location; the actual effort, however, depends on the prepared assets, the necessary tests, and pre-production. Reshoots too can potentially become easier and more plannable because the digital environment is preserved and thus supports continuity.

Fraser offers a concrete example for twilight scenes. For the construction site overlooking Gotham in “The Batman,” the volume worked “exceptionally well” because the light was soft and one otherwise only has a short real window for twilight shots (Hollywood Reporter). “House of the Dragon” director Clare Kilner calls the method “particularly beneficial for endless sunrises and sunsets,” but concedes that it worked well for some scenes and less well for others (Hollywood Reporter).

Because the outer frustum turns the wall into a light and reflection source, realistic reflections and light interaction arise directly on performers and props instead of being laboriously added in post (Epic Games). There are ecological arguments too, though they must be placed carefully in context. A study by Filmakademie Baden-Württemberg measured, in a concrete test comparison of two student productions, 1,594 kWh for virtual production versus 5,073 kWh for a comparable offline-rendered production, roughly one third of the energy demand (Filmakademie Baden-Württemberg, Sustainability Report). These figures are explicitly a single test case under limited assumptions, not a general rate for all virtual productions, and not a direct overall comparison between shooting at a real location and shooting in the volume: virtual production was compared against an offline-rendered production, not location versus studio. The study also excludes energy for cooling and data storage and stresses that the advantage applies above all when renewable energy is used (Filmakademie). For the broader context, it cites an average output of 2,840 tons of CO2 for large tentpole productions, but itself points to the narrow limits of its measurement.

The limits: physics, aesthetics, and the danger of uniformity

As great as the control is, the limits are just as clear. Physically, the wall remains a wall: for a fully virtual environment, Epic recommends a volume of at least 270 degrees, and the finite surface limits depth and camera movement (Epic Games). From a cinematographer's point of view, the frustum must always stay in front of the camera; very fast or large movements can look beyond its edge (Frame.io Insider). Then there is moiré, an interference pattern that arises when display and sensor pixels are slightly offset; it appears more strongly when the focal plane lies on the LED or the camera is angled to the wall (Epic Games). Visible pixels and scanlines remain a risk that is mitigated with finer pixel pitch and higher refresh rates. Manufacturer Sony specifies typical pixel pitches of 1 to 4 millimeters for volumes and recommends 1.25 to 2.5 millimeters for small stages up to about eight meters wide, so that individual pixels do not become visible – a vendor claim (Sony, Crystal LED White Paper).

There are hard limits when it comes to light as well. Fraser says bluntly that the volume struggles to reproduce midday and daylight convincingly; bright desert sun at noon is “very hard to do,” and pulling the level of the sun from LEDs is “impossible” (Filmmaker Magazine). From a cinematographer's perspective, the color quality of the light emitted by RGB LED walls is viewed critically (Frame.io Insider). Add to this the economic hurdles: VFX expert Ben Grossmann estimates the construction cost of an LED stage at 3 to 30 million US dollars, with the digital content often the largest item and full CG environments potentially requiring four to six months of work; he even sees an oversupply of stages that many do not yet know how to operate (Hollywood Reporter). Because creative decisions have to be made earlier, short-term flexibility decreases, and job profiles shift toward the Virtual Art Department. The decline of classic shooting days is measurable: in Los Angeles, on-location shooting days fell to 19,694 in 2025, a drop of 16.1 percent compared to 2024, with FilmLA seeing the cause primarily in global competition and cost issues, not in virtual production alone (FilmLA, 2025 Report). The real aesthetic danger is uniformity: when efficiency, control, and cost savings become the dominant design criteria, sterile images threaten, images that lack the natural disorder, the local textures, and the cultural details of real places.

Why a real place sometimes becomes a character in its own right

A real location is more than a backdrop. It provides natural light, real weather, spatial constraints, architecture, local sounds, and surfaces; it produces chance encounters and cultural authenticity and influences acting and directing. It is precisely this unpredictability that a studio cannot automatically create. Fraser puts it succinctly: “If you're in the right location at the right time, you shouldn't need much light” (Filmmaker Magazine). And Jay Holben describes the lasting value of place: there is nothing comparable to standing in a real canyon or visiting an actual castle in Ireland; practical locations offer “a wealth of discovery” (Hollywood Reporter).

This is precisely where the central risk of misuse lies. Fraser warns against the misconception that an LED volume solves all the logistical problems of location shooting; the risk lies in not understanding what it is suited for and what it is not – scenes that do not belong in the volume lead to weak images (Hollywood Reporter). An anonymous industry insider sums up the divide drastically: some productions have succeeded brilliantly, others have failed due to a lack of preparation (Hollywood Reporter).

Counterarguments, limits, and risks: the honest reckoning

Anyone who rejects modern technology across the board makes it too easy for themselves. The volume has real, demonstrable strengths, and there are scenes for which it is superior to real places. But the balance sheet has a downside. First, the ecological calculation is incomplete: the Filmakademie study excludes cooling and data storage and stresses that virtual production is “not a solution for all aspects of a film production” (Filmakademie). Second, vendor promises must be read with caution: Sony puts the mirror reflection of its B series at a DOI value of “0” and advertises anti-reflective coating and a high frame rate against scanlines – all manufacturer claims that must be tested in everyday production (Sony White Paper). Third, the technology shifts labor and value creation: it can displace local crews and service providers at classic locations, while new roles emerge in pre-production.

The most convincing answer is therefore not an either-or decision, but the hybrid solution. Real locations, physical sets, practical effects, miniatures, LED volumes, green screen, classic VFX, and AI-assisted post-production can be combined. From a cinematographer's point of view, rear projection, live action, and miniatures still work “just as well as they always have,” and real places can be scanned in via photogrammetry and reused in the volume (Frame.io Insider). Netflix bundles precisely these building blocks in Eyeline, but explicitly classifies generative AI as a supplementary building block alongside virtual production and classic VFX, not as their replacement (Netflix, Eyeline).

The conclusion is therefore this: a studio can simulate almost any place, but it does not automatically create that unpredictability through which a real location sometimes becomes a character in the film itself. The control of the volume is a tool, not a goal. Whoever uses it because a twilight would otherwise have only a short window gains from it. Whoever uses it to replace desert sun, real weather, and the random texture of a real place risks images that look flawless and yet feel empty. The future does not belong to the box versus the place, but to the smart decision about when which space tells the better story.