Spotlight on Web3 Productions: inside Dubai’s first AI-native virtual production studio

What if a production could arrive at a studio with a single reference image and be shooting inside that world the same afternoon?

That is the thinking behind Web3 Productions. Based in Dubai Production City, the studio opened at the end of 2025 after around two years in development. Today, it operates two Disguise-powered stages, a dedicated render farm and a team spanning virtual production, broadcast, Unreal Engine and generative AI.

We spoke to Tazio Simoncioni, CTO at Web3 Productions, and Tristam Thomas, Chief Creative Officer of the company’s AI production arm, about building a workflow designed around speed, flexibility and keeping creative decisions in human hands.

Building a studio around speed

For Web3 Productions, the starting point was a simple challenge. Virtual production had already made it possible to place talent inside almost any world, but creating those worlds could still take weeks.

“Content was still taking weeks of design before a single frame reached the wall,” explains Simoncioni. “That gap is where projects die, because the client changes their mind on day three and the schedule cannot absorb it.”

The studio was designed to close that gap. Its team includes Unreal Engine artists working across modelling, texturing and environment creation, alongside creative technologists specialising in generative workflows. Broadcast and engineering specialists, including a certified Disguise trainer, complete the team.

“These are new crafts, not shortcuts,” says Simoncioni. “Someone still has to know what a good frame looks like, what the light should be doing, why a lens choice matters. The tools moved the starting line. They did not remove the people who have to cross it.”

From one image to a filmable world

At the heart of the workflow is an application developed in-house, known internally as ALBA.

A production can begin with a single reference image. The image is refined and expanded into a 360-degree panorama before being developed into a 3D Gaussian splat, creating an environment with real depth rather than a flat background.

Volinga then brings that environment into Unreal Engine, where it can be used as a filmable set and delivered to the LED volume through RenderStream.

The application is connected directly to both Unreal and Disguise through their APIs, helping remove manual steps between creating content and getting it onto the stage.

“Show us one picture. A few minutes later the crew is standing inside it, and the conversation has already moved on to lighting and performance,” says Simoncioni.

For locations that need to be recreated precisely, Web3 Productions can also use drone footage and 360-degree video to build photoreal environments. Throughout the workflow, colour is managed using OCIO transforms through the Disguise VP plugin, helping keep what is generated, rendered and ultimately captured on camera consistent.

Speed is only valuable if filmmakers can still make creative decisions. Every environment created by Web3 Productions exposes parameters in Unreal that can be carried through to Disguise. If a client wants to change a scene from midday to golden hour, for example, the team can adjust the lighting without regenerating the entire environment.

Twenty-two worlds, one stage

One of the largest productions currently using the facility is an unannounced vertical micro-drama series spanning 61 episodes and 22 environments.

Twenty of those environments run on Web3 Productions’ main stage, with only two built as practical sets. Most of the virtual environments were built directly as assets in Unreal by the studio’s artists. Around 30% followed a different route: the hero image for the environment was created and edited in Disguise’s Nano Banana plugin, then the set was reconstructed in Marble or Spaitial as a 3D Gaussian splat and brought into Unreal from there.

Nano Banana allowed the team to make targeted changes to the approved hero image – such as removing signage, changing a floor or moving a prop – without regenerating the rest of the frame, helping ensure the environment that reached the wall remained the one that had been signed off.

“We chose per environment, not per ideology,” says Simoncioni. “Some sets want the speed and the photoreal depth you only get from a splat. Others need an art director to place every object, and for those our artists build in Unreal from nothing. The value is being able to choose.”

The production is shot at 25fps from 6K RED originals and finished natively in ACES. Across the series, the Disguise setup helps the team maintain the same colour transforms, mapping and environments from shot to shot.

“A series this size would normally mean scouting, permits, build days and travel for twenty-two sets,” Simoncioni adds. “We are doing it on one stage, and the world changes between setups.”

Powering the workflow with Disguise

Web3 Productions runs two independent Disguise setups.

Stage A uses four VX 4+ systems alongside eight RX III render nodes. Studio B, designed for broadcast, has two VX 2+ systems and two RX III render nodes, as well as three tracked PTZ cameras and a full production gallery.

A separate farm of eight RTX A6000 servers supports generative content creation, helping ensure those workloads do not compete with live production.

Across the facility, Disguise’s Designer software maps and controls the LED environments, RenderStream connects Unreal to the volume, the VP plugin supports colour management and Shot Recorder captures camera data for downstream workflows.

For Simoncioni, the openness of the platform was key. “Our pipeline depends on pushing finished worlds into the servers automatically,” he explains. “RenderStream plus an API surface we could build against made Disguise the only platform we could treat as a programmable endpoint rather than a closed media server.”

The same setup also supports live production. Multiple RED cameras can be switched frame by frame while the perspective rendered on the LED follows the active camera, allowing Web3 Productions to combine cinematic image quality with multi-camera broadcast workflows.

Story first

Alongside its virtual production business, Web3 Productions operates a dedicated AI production arm exploring virtual practical, hybrid and fully AI-generated production workflows.

For Thomas, however, the technology remains secondary to the reason for using it. “Story is at the heart of everything we do,” he says. “We treat AI as a visual effects tool, and the point of that tool is to extend the human capacity to tell and to appreciate stories.”

It reflects a broader philosophy across the studio: rather than choosing between virtual production, AI or traditional filmmaking, Web3 Productions wants to give creators the freedom to choose the right approach for every shot.

What comes next

The team is now looking at how more of the preparation process can happen before a crew arrives on set, with previsualisation, prelight and test renders generated from a script overnight.

For Simoncioni, it is another step towards a future where virtual production, AI and traditional filmmaking become part of the same flexible production pipeline.

“The market still talks about these as separate species,” he says. “They are going to merge. There will be one pipeline, and one studio decides per shot which version of it is right, because the honest answer changes with the schedule, the budget and what the director actually wants.”

“The job is not to pick a camp. It is to be able to choose, on the day, and be right.”