Gengis AI is bringing production-ready workflows to Asia’s virtual sets

By Shaun Lim

Artificial intelligence (AI) is reshaping media, but the gap between promise and practical deployment remains wide. For most production teams, the challenge is not imagination; it is execution under real-world constraints.

AI promises transformation for media, yet today most teams struggle to deploy it safely and effectively on-set, said Joel Lim, Co-Founder & CEO, Gengis AI, a production-ready AI-enabled studio based in Singapore. 

“As a production partner, Gengis AI is embedded in the planning, not bolted on at the end,” Lim told APB+. “We’re involved at script breakdown, blocking, camera decisions, environment planning, and post assumptions. This means the AI workflows are designed around what the crew can execute, not what looks impressive in isolation.”

A virtual production specialist, Lim decided to join forces with veteran producers and filmmakers Karen Seah and Tian Sian Ju, the co-founders of Refinery Media, to launch Gengis AI earlier this year, with the aim of designing and operating AI-assisted workflows that integrate across development, scripting, on-set virtual production, and post production.

Instead of being just a technology partner, Gengis AI is positioning itself as a production partner, one that is not afraid to ask uncomfortable questions.

Lim illustrated, “What can realistically be locked on-set? What needs flexibility in post? Where does AI reduce risk, and where does it introduce risk?

“If something breaks on shoot day, it’s our problem too. That mindset changes how conservative and disciplined you are about using AI at all.”

As starting points, Gengis AI is focusing on two high-impact areas: AI-driven VFX generation for rapid creation and enhancement of digital assets, and intelligent virtual set extensions that optimise LED walls for dynamic, real-time environments.

As these capabilities continue to be enhanced, which parts of the filmmaking and virtual production pipeline does Gengis AI see the most immediate and tangible gains from deploying AI?

“The most immediate gains are coming from areas where AI can be integrated into proven production pipelines, rather than treated as a standalone replacement,” said Lim.

Specifically, he sees environment creation and set extension as seeing the fastest and most reliable returns. “AI lets us move faster and explore scale earlier, but final delivery still runs through traditional VFX processes. We composite, colour match, and manage texture fidelity, grain, and resolution using established pipelines. In many cases, AI is deliberately invisible, embedded inside workflows audiences already trust.”

When it comes to performance and long-form continuity, Gengis AI takes a more selective approach. While they are actively testing and releasing work from fully AI-generated video, Lim was quick to remind that context matters. While short-form or concept-driven pieces can tolerate more abstraction, narrative and performance-led work still benefits from being anchored in reality, he explained.

“Our principle is to root critical elements first, then extend outward with AI where it holds up under scrutiny. As the technology matures, those boundaries evolve. The goal isn’t to avoid fully AI work, but to use it intentionally and let production standards, not novelty, determine how far it goes.”

Another area of focus for Gengis AI is to ensure that human creators remain the core of the creativity process. To decide where AI adds value, and where it should be omitted altogether, Gengis AI designs guardrails defined not by ideology, but by outcomes.

The first guardrail of control looks at how AI deliberately shapes the result, iterates towards a specific creative intent, or explains why something looks the way it does. If it fulfills none of those functions, AI is removed from that part of the pipeline.

The second guardrail of integration examines how AI can slot cleanly into existing production and post workflows. If AI forces compromise in colour, finishing, or delivery downstream, it will not be used.

The third, and arguably most important guardrail, is audience tolerance. While short-form or concept-driven work can absorb more abstraction, narrative, performance-led, or broadcast work has a much lower tolerance for inconsistency or artefacts. 

Lim highlighted, “If AI doesn’t meet those criteria, we don’t force it in. The goal isn’t to maximise AI usage, but to deliver work that feels intentional, coherent, and finished. As the technology improves, the boundaries move, but the principles stay the same.”

Having already started operation, one of the first projects undertaken by Gengis Ai is an unscripted, vertical reimagining of Refinery Media’s established SupermodelMe IP. The project, says Gengis AI, serves as a live production showcase on how AI can compress development timelines and expand visual scope while maintaining operational efficiency.

Lim elaborated, “For most narrative or performance-led work, changing format still means rethinking blocking, pacing, and coverage. That’s a creative decision, not a technical one, and AI doesn’t remove that work. In cases like SupermodelMe, reshooting is still the right choice”

Where AI provides gains today, is in language and localisation. Particularly, he sees dubbing, subtitling, and timing adjustments improving quickly, creating real impact for regional distribution in Asia.

Beyond that, the benefits are more incremental, Lim said. For instance, AI can help explore options faster, visualise alternate framing, or extend environments, but it does not replace intentional design for a platform. “Its real value is lowering the friction to test and validate decisions earlier, so when you commit to a new format or region, you do it with more confidence and less waste,” he added.

While AI is currently being applied more in VFX and virtual production, it is also slowly creeping into the realm of live and near-live workflows, where little room for error exists.

Before looking at helping address the challenges of working in real-life environments, Lim was quick to clarify, “When people talk about live or near-live workflows in virtual production, they usually mean visuals that need to behave predictably in real time on set, not AI generating imagery live during capture.”

In that context, Gengis AI is ready because they place AI upstream of the real-time moment, not inside it. AI is used to design, generate, and validate environments, set extensions, and looks before the shoot, under the exact camera, lens, resolution, and colour constraints the production will use. By the time content is rendered on an LED wall, it is stable and deterministic and behaves like any conventional real-time asset.

“What we avoid is introducing generative variability during capture,” said Lim. “Anything that affects performance, timing, or continuity needs to be locked before cameras roll.”

The remaining challenge, he revealed, is observability and control. “Until AI systems can be monitored and constrained with the same confidence as other production tools, their role in live environments should remain preparatory, not improvisational.

“Used this way, AI makes real-time workflows more robust by reducing uncertainty before the shoot, where it belongs.”

As Asia’s production ecosystem accelerates, the role of AI will keep shifting from experimental add-on to embedded discipline. 

For Gengis AI, the future is not about replacing filmmakers and content producers. Instead, it is about giving them sharper tools, faster iterations, and more confidence to scale ambition across formats and regions. 

The next chapter of media innovation will be written not by AI alone, but by how human creators choose to wield it.