Atelico opens on-device AI engine in Early Access for studios

Amara Osei

Studio Atelico has launched the Atelico AI Engine into Early Access with on-device image generation, language models, speech systems and AI-driven character behaviors across Windows, macOS, iOS and Linux.

What can studios run locally?

The engine works with Unity, Unreal Engine and Godot. Studio Atelico says developers can run AI workloads beside gameplay without affecting rendering performance, using either GPU or CPU and removing cloud dependencies.

Its announced toolset covers:

• image generation;
• language-model execution;
• speech systems; and
• AI-driven character behaviors.

Studio Atelico first used the technology in Bobium Brawlers, its debut game, which the studio designed to showcase the engine. For studios, the immediate proposition is compatibility: they can test local AI alongside three established engines rather than move the whole project onto a new stack.

Why does local execution matter?\n

CEO Piero Molino argues that developers have faced a false choice between intelligent gameplay, performance and creative control. He said:

"We built the Atelico AI Engine so studios don't have to make those trade-offs. The studios that succeed won't be the ones using the most AI, they'll be the ones using AI entirely in service of the player experience."

That is Atelico's product claim rather than an independently reported benchmark. The announcement provides no frame-time data, minimum hardware specification or performance comparison, so studios evaluating the engine will have to do the less glamorous work of testing it on their own rigs.

What does early access change?

Developers can request access through Studio Atelico's website, and that request is the only access route described. The announcement names no price, licensing structure or general-availability date. That makes Early Access a controlled introduction, not a fully specified commercial release.

In an earlier GamesIndustry.biz interview this year, Molino framed the long-term goal as becoming the Epic Games of AI. He described a self-reinforcing loop in which new technology enables new games and the studio's work on those games feeds back into the technology. For a platform company, that means the engine needs both developer adoption and a compelling showcase; Bobium Brawlers currently supplies the latter.

How is artist training governed?\n

Studio Atelico's AI-first ecosystem trains generative models on work from its artists. Chief product officer Chris Olson said:

"We're not looking to reduce steps in the workflow so that you can fire half your art team. We have an art director. We pay our artists."

Olson said the company recently released a draft document called ARC, the Artist Rights Contract, intended to protect an artist's livelihood and art. It allows an artist to define what their art can be used for, participate in the upside when that art trains models, and specify individual use cases.

Olson said ARC currently applies only to Bobium Brawlers; using the art for other purposes requires renegotiation.

Molino said Atelico trains its own models with its own art and in-house-created decks, which means the studio owns the underlying data. He described that arrangement as ethical and said it does not replace staff. Instead, he said, the models enable projects that were not possible before.

What should studios watch?\n

The immediate test is operational: can local AI coexist with rendering across four operating systems and three engines while keeping the artist terms that Atelico says guide the system? The next signal worth watching is whether ARC expands beyond Bobium Brawlers, because every new use currently requires a fresh negotiation.