adminOn-device generative AI is central to Studio Atelico’s approach to building games that combine player creativity with traditional design. The studio’s debut title, Bobium Brawlers, uses its own AI technology to create creatures, while the company says it aims to protect artists’ rights and avoid relying on expensive cloud computing.
Image credit: Studio Atelico
Studio Atelico introduced its Atelico AI Engine in January last year. The technology is designed to run on players’ devices rather than sending every request to remote data centres. Six months later, the studio raised $5 million in seed funding to develop its work in generative AI for games.
The team is now building Bobium Brawlers, a turn-based creature-battling game that also serves as a practical demonstration of the engine. Its creature-generation models are trained on artwork made by the studio’s own artists.
On-device generative AI and artist rights
Studio Atelico chief product officer Chris Olson acknowledges that the use of generative AI raises ethical questions. He says the studio wants its technology to support creative teams rather than reduce the need for artists. The company employs an art director and pays its artists for their work.
Studio Atelico has also shared a draft framework called the Artist Rights Contract, or ARC. It is intended to give artists a say in how their work is used and a chance to benefit when it contributes to AI models. Permissions are tied to specific uses: using artwork for a different project would require a new agreement.
CEO Piero Molino says the studio trains its models using material it owns, including in-house artwork and card designs. For developers assessing similar systems, the practical takeaway is to establish clear permissions and define the intended use of training data before a model is built.
Why run AI on the device?
At Nordic Game, co-founder and game design director Ennio De Nucci described the engine as a code library that developers can integrate into different kinds of games. Because it runs locally, it can reduce dependence on cloud services and does not require a continuous internet connection for every generation request.
De Nucci also argues that a focused task, such as creating creatures in a particular art style, may not require a vast general-purpose model. A smaller, specialised system can be a better fit for a specific game feature. It can also help developers avoid recurring charges for cloud-based AI services.
For mobile developers, the choice between local and cloud processing depends on the feature, device capabilities and the player experience. On-device processing can offer offline access and predictable service costs, while teams still need to consider performance and the range of devices they support.
Bobium Brawlers uses AI to generate creatures. Image credit: Studio Atelico
Combining AI with familiar game design
Molino believes that players may feel a distinct connection to creatures they create through personalised prompts. Rather than choosing only from a fixed set of characters, players can describe an idea and see a creature appear in the game.
The studio says it explored around ten prototypes and mechanics while considering how AI could shape a game. Its view is that successful AI games do not have to abandon established design. Bobium Brawlers aims to combine structured, familiar gameplay with the more open-ended possibilities of creature creation.
This balance is useful for developers to consider: AI features should serve a clear gameplay purpose, not simply be added because the technology is available. Studio Atelico describes the title as relatively modest in scope, reflecting that there is no settled formula yet for designing AI-based games.
Scaling the Atelico AI Engine
The $5 million seed round gives the studio backing as it develops both its game and its technology. Olson says the distinctiveness of the project could help the company identify its audience, while Molino argues that a differentiated game may need a different approach to finding players than a familiar mobile genre.
Bobium Brawlers is planned as a free-to-play title. Since its AI engine runs on-device, the studio says it can avoid building its business around subscriptions intended to cover ongoing cloud-generation costs.
The game is also a showcase for the Atelico AI Engine. The studio’s longer-term aim is to make the technology more flexible, so developers can combine AI components for their projects instead of depending on fixed cloud-based tools. Studio Atelico plans to make smaller games, learn from player feedback and refine its approach; its stated target is about one game per year, depending on each project’s success.
“Our long-term aim would be to be the Epic Games of AI”
Piero Molino
Molino compares this gradual development process with the evolution of early first-person shooters, where experimentation helped developers discover what made the format work. Studio Atelico hopes to follow a similar cycle: build with new technology, release games, gather feedback and improve future projects.
What on-device generative AI could mean for games
Studio Atelico’s approach brings together local processing, artist-focused policies and AI-driven player creativity. Its plans are still developing, but Bobium Brawlers offers a concrete example of how generative tools can be designed around a specific game mechanic rather than used as a general feature.
For studios exploring on-device generative AI, the key considerations are clear data permissions, a well-defined use case, device performance and a gameplay experience that remains enjoyable without the AI feature alone. These principles are central to Studio Atelico’s vision for its engine and its first game.
