The Divinity Developer Explains Its Application of Machine Learning for Upcoming Project

The studio behind acclaimed titles like Baldur's Gate 3 and Divinity: Original Sin has recently teased its upcoming project, creating a wave of anticipation within the gaming community. However, follow-up remarks from the studio's co-founder have introduced a new dimension to the discussion, focusing on the developer's approach toward AI tools.

A Tool for Ideation, Not Replacement

In a recent message, Larian's director explained that the team is employing generative AI for certain supporting tasks. These involve fleshing out pitch decks, generating rough visual ideas, and writing placeholder text.

Importantly, Vincke stressed that the end material in the game will be created entirely by real creatives. "Larian is creating every line in-house," he affirmed.

Larian is constantly increasing our team of storytellers and are busily assembling writing teams.

As visual development is being particularly referenced — we presently have twenty-three concept artists and have roles to fill for further talent.

Each initiative we do is incremental and designed to letting our team spend additional energy on the creative process.

Every machine learning application implemented properly is supplementary to a artist's routine, not a replacement for their talent.

Addressing Concerns and Clarifying the Vision

The revelation of using AI originally generated backlash among a segment of the community. In reaction, Vincke provided further detail on public forums.

"At Larian, we employ machine learning to explore references, in the same way we use Google and art books," he stated. "During the initial ideation stages we use it as a simple sketch for composition which we then replace with hand-crafted illustrations."

He continued, "Our studio recruits creatives for their creative vision, not for their ability to execute what a AI generates."

Focused Uses for Machine Learning

Vincke had earlier outlined the company's targeted approach to machine learning, defining its use into key functions:

• Handling Monotonous Jobs: This encompasses motion capture cleaning, dialogue cleanup, and Larian-specific work like retargeting animations.
• Accelerated Iteration: Using technology to rapidly prototype simple mock-ups of gameplay ideas to experiment with concepts ahead of full development.
• Future Potential for Gameplay: Exploring how machine learning could in the future enhance emergent reactivity, especially in simulating player-driven narratives in a vast role-playing world.

He clearly stated that key artistic disciplines — such as writing — are not areas where the team is reducing human talent. On the contrary, Larian is expanding its staff in these exact roles.

"Larian is neither shipping a game with machine-made assets, nor considering cutting teams to substitute them with AI," Vincke summarized.