Choose LED, green screen, or plates one shot at a time

Start with the shot, not the technology

An LED volume, a keyed backing, and a photographed plate solve different problems. They can all appear in the same sequence. The mistake is asking which method is best for a whole show, then forcing every setup through that answer. Instead, write a row for every distinct camera setup or repeatable setup family: the subject, lens range, camera movement, physical set edge, interactive light requirement, final background need, and post tolerance.

Epic’s ICVFX best-practices documentation makes a useful, limited point: an environment that looks acceptable on a workstation may not work on an LED wall, and it recommends testing on equivalent stage hardware. That does not prove a wall is right for a particular shot. It does mean a treatment should not promise final pixels solely from a desktop preview.

Use four questions in order

First, does the actor or practical object need to receive the background’s moving light, color, or reflections? If yes, an LED wall or a separately planned interactive-light solution deserves serious consideration. Second, does the camera see a shallow, hard-to-key edge such as hair, transparent material, smoke, or a moving reflective surface? That raises the cost and uncertainty of a green-screen composite, but it does not automatically rule it out.

Third, can the required background be captured as a stable plate with a camera move that is known and repeatable? A plate can be the simpler answer when the foreground does not need to react to a live rendered world. Fourth, where must the crew make a decision? If the director needs a horizon, parallax cue, or eyeline in the take, an LED approach can move that decision earlier. If the image must stay deliberately open for later redesign, keying may preserve more latitude at the price of later work.

Make a provisional choice visible

Atlas proposed process: assign each row a primary method, a fallback, and one failure condition. The primary method is the current plan; the fallback is what the schedule can absorb if the test fails; the failure condition is observable. “Wall looks wrong” is not usable. “The 40 mm close shot produces unstable fine fabric against the background at the approved stop” is.

Epic’s quick start describes a cluster, tracked camera, and inner/outer-frustum setup as components of an ICVFX workflow. Those are implementation facts, not a mandate to build that configuration for every background shot. Keep that distinction in the treatment.

Illustrative example

Illustrative: a dialogue scene has three setups. A window-side close-up needs sunset color across an actor’s face and a glass tabletop; the team selects LED as the first test. An over-the-shoulder toward a soft exterior needs no usable reflection; green screen is selected, with an interactive-light cue specified separately. A locked insert of a phone on the table uses a supplied exterior plate because neither tracking nor actor eyelines are involved. The result is not a hybrid workflow for its own sake. It is three decisions with different evidence.

• List the required lens, movement, focus behavior, and visible physical/virtual seam for each setup.
• State what must be present in camera and what may remain a post decision.
• Test the riskiest shot, not the most attractive still frame.
• Record the fallback and the deadline for choosing it.