On the road to text-to-videogame. Why we invested in Ludus AI.
In the era of generative AI—when entire industries have already been reshaped by text-to-speech, text-to-code, text-to-image, and text-to-video models—a natural question arises: when will we see true text-to-videogame? When will a model—and a startup—emerge that enables game creation to be radically simplified, in the same way images or app prototypes are created today?
This question is not purely technological. It is primarily economic. By various estimates, the global gaming market is now 2–3 times larger than the music market and significantly larger than the global film box office. Moreover, gaming is growing faster, engages billions of users, and—unlike film or music—is built on interactive systems rather than linear content. It is precisely this systemic nature that makes the potential for automation and generativity in games theoretically enormous.
Indeed, AI tools have already found a strong foothold in game development. Models generating 2D and 3D assets, supporting animation, textures, audio, or gameplay testing are becoming increasingly common. At the level of the “production environment,” AI is meaningfully reducing costs and shortening development cycles.
There is, however, one area that remains a clear bottleneck: game engines.
Game engines as the bottleneck for AI adoption
The game-engine market is highly concentrated. For years, a small number of solutions—primarily Unity and Unreal Engine—have dominated, powering the vast majority of global 3D game productions. Industry estimates suggest that Unreal Engine accounts for roughly 30–40% of the game-engine market, with a significantly higher share in AAA productions and highly complex projects. Unreal drives a large portion of the biggest game releases, as well as a growing number of projects outside traditional game development—from VFX and film, to automotive, simulations, and digital twins.
Crucially, Unreal Engine is currently seeing the fastest adoption growth among available engines. As this expansion continues, so does project complexity: scale, lifespan, technological dependencies, and maintenance costs all increase.
The paradox is that while AI performs well on the periphery of the creative process, it struggles to penetrate the very core of engine-level work. General-purpose copilots and language models:
- do not understand Blueprints as logic graphs,
- do not operate on the context of an entire project,
- hallucinate solutions incompatible with specific engine versions,
- and do not integrate with real-world team workflows.
As a result, AI remains a tool “next to production,” rather than an integral part of it. For enterprise teams, there is an additional barrier: IP security, NDAs, and the lack of consent to process proprietary code in public models.
If text-to-videogame is ever to become a reality, the first step is therefore not generating entire games, but solving a far more grounded problem: how to manage the complexity of modern projects built inside game engines.
Before “creating games,” first “understanding games”
Modern games—and, more broadly, interactive 3D experiences—are not created from scratch in a single creative cycle. They are built over years, by teams ranging from a few to several hundred people, on projects that evolve, change direction, and accumulate technical debt. In such an environment, the biggest challenge is not creating a new game, but maintaining, evolving, and understanding the one that already exists.
Blueprints in Unreal Engine illustrate this problem perfectly. On one hand, they allow teams to build game logic quickly in a visual way. On the other, in larger projects they become difficult to audit, refactor, and transfer to new team members. Knowledge of “how the project works” often exists only in the heads of a few senior developers. This is not an issue of code aesthetics, but of real operational risk: bugs, performance degradation, and blocked further development.
If we therefore treat text-to-videogame not as a single technological leap, but as a gradual process of pushing the boundary of what AI can do inside an engine, the first logical step becomes clear: tools that help teams understand, organize, and maintain existing projects.
This is where the Ludus AI story begins.
What Ludus AI is and the value it delivers today
Ludus AI does not try to replace the game engine or promise full automation of the creative process. Its starting point is far more pragmatic. As Unreal Engine becomes the standard for an increasing number of use cases—and its complexity grows faster than the supply of experienced specialists—the greatest value lies not in generating new code, but in reducing the cost of understanding and maintaining what already exists.
At the core of the product is deep, project-specific context. Ludus analyzes Blueprints, related C++ code, assets, and the specific engine version a team is working on. Key use cases include:
- Blueprint analysis and refactoring,
- project quality audits and identification of technical debt,
- rapid onboarding of new team members,
- support for teams maintaining long-lived projects.
In conversations with users and studios, Blueprint analysis and overall project understanding most frequently emerge as features that are difficult to replace with any other AI tool.
The second pillar is enterprise-grade architecture. From day one, Ludus has been built for customers who:
- cannot move code outside their own infrastructure,
- require full control over data and costs,
- work on custom versions of Unreal Engine.
Private-cloud and self-hosted deployments make the tool deployable in high-security environments—which, in practice, is a prerequisite for adoption in larger organizations.
Adoption and where the company is today
At the time of investment, Ludus AI shows clear bottom-up traction:
- over 25,000 registered users,
- more than 1,000 monthly active users,
- several hundred paying subscribers,
- strong month-over-month revenue growth.
This adoption has been largely organic. While organic growth is hard to sustain indefinitely, it is an excellent source of data, feedback, and rapid product iteration. In parallel, an enterprise path is being developed—not as a deck slide, but as a real product direction, with infrastructure already being built to support it.
Beyond today: text-to-videogame as an industry direction
Ludus AI has an ambitious long-term vision: gradually moving toward a world in which AI takes over an ever-larger share of the creative process—from logic, through assets, to complete interactive experiences. In other words: toward text-to-videogame.
We believe this is a real direction for the entire industry, not a marketing fantasy. It is already visible in the growth of UGC, procedural design, and the increasing role of AI in production pipelines. At the same time, we are very aware that the path to this goal is long.
That is why our investment thesis is different: before AI starts “creating games,” it must first become part of everyday workflows inside engines. It must understand projects, their history, and their constraints. It must help teams work faster and more safely here and now.
In this sense, Ludus is not a shortcut to the future, but one of the necessary steps toward it. First a niche. First a copilot standard for Unreal Engine. Only then—optionally—broader generativity.
Why we invested
We invested in Ludus AI because we see a rare combination of:
- a large and growing market,
- a clear, structural problem,
- a specialized product that is hard to copy,
- and a long-term vision aligned with the direction of the entire industry.
And above all, a founder who lives and breathes the product.
The risks are real: dependence on the Unreal ecosystem, competition, long enterprise sales cycles. At the same time, they are addressable—as long as the company maintains discipline and focus on solving a very concrete problem.
Our investment is not based on the assumption that Ludus will “build a new game engine.” It is based on the conviction that whoever wins the tooling layer around Unreal Engine gains a privileged position to participate in the next evolution of game creation and 3D worlds.
How this competition will ultimately play out remains uncertain. We may underestimate the pace of generative-model development; perhaps models like GameFactory or NVIDIA Cosmos will learn physics simply by observing other games, and classical engines will no longer be needed. What is certain, however, is that the industry is in a phase of dynamic AI adoption—and we believe Ludus has an excellent starting point to compete for becoming one of the new standards and reference points.