Synapse WorldGen™

Sourcing content for the LED volume is one of the biggest hurdles in virtual production. Synapse WorldGen uses a bespoke hybrid VP/AI workflow to quickly generate detailed environments for the volume, every one guided by real-world photography so it stays photoreal and art-directable. It covers both halves of the job: frame and light the shoot with a generated environment on the wall, then replace or finish it in post with the performance, lighting and camera move carried through from the day.

Shooting with a generated environment on the wall

The hardest scheduling problem in virtual production is that the environment has to be finished before the camera rolls. WorldGen breaks that dependency. A generated environment goes on the wall for the shoot, which is enough to give correct framing, eyelines and real interactive lighting on talent and set. You capture the plate you need without locking the final look while the creative conversation is still open.

You capture the plate you need without locking the final look.

Why WorldGen starts from real photography

We don't do simple text-to-image prompting. Every WorldGen environment is anchored by real-world photographic elements, sprites and lighting and texture references that guide the final output, not only for maximum photorealism but for artistic control and scene continuity across a sequence.

Replacing the environment in post

After the shoot, the on-volume environment can be replaced entirely. Because the light in the room came from the wall, the replacement inherits a lighting relationship that was physically real: the actor was lit by something the right color, from the right direction, at the right intensity. That is what separates this from a green-screen composite, where the light has to be invented and rarely convinces.

AI Gaussian splats: generated worlds in 3D

WorldGen isn't limited to static generated background plates. It can also deliver full 3D assets that allow dynamic, real-time camera tracking and parallax in camera. WorldGen environments are essentially splatted into a 3D asset, imported into Unreal Engine and used like any other tracked 3D VP environment.

AI finishing for LED-wall scale

The same pass finishes shots that are not holding up. Soft or lower-resolution content that falls apart at volume scale gets lifted to photoreal, tuned specifically for the demands of an LED wall where scale is unforgiving. Frames captured off the volume, including those that see past the edge of the wall, run through the same workflow.

This is AI-assisted finishing rather than AI replacement. Every result is art directed and taken to the same standard as the rest of the volume content, because a generated frame that does not match the show is a note, not a delivery.

What WorldGen covers

• Generated environments: detailed worlds for the volume, anchored by real photographic references
• AI Gaussian Splats: generated environments delivered as tracked 3D assets in Unreal Engine
• Temp to final: a generated environment on the wall for framing and light, replaced or finished in post
• AI finishing: lower-resolution content lifted to photoreal for volume scale
• Art directed throughout: output held to the show's standard, not a prompt result

AssetScreen vs. green screen

Green screen also defers the environment decision, and it costs you everything the wall provides. Even when an AI backdrop is only a temp background, it keeps the backbone of an LED stage:

• Filmmakers have an environment, a proscenium to frame against.
• Every department sees the comp live and contributes to one vision.
• Light and reflections from the LED environment land on the set and on the talent's skin.
• If the backgrounds are replaced in post with plussed-up assets, keying becomes simple hold-out mattes with interactive roto edges, because the background already carries the same tones, textures and contrast.

The set is never green, and the light on the day is never a guess.