The Divinity Developer Explains Its Application of Generative AI for Upcoming Project

The developer behind acclaimed RPGs like Baldur's Gate 3 and Divinity: Original Sin recently teased its new project, generating a wave of anticipation within the player base. However, recent remarks from the studio's co-founder have added nuance to the narrative, focusing on the team's philosophy toward machine learning.

AI as a Creative Assistant, Not a Substitute

In a recent clarification, Swen Vincke detailed that the team is utilizing machine learning for certain supporting purposes. These encompass fleshing out PowerPoint slides, producing initial visual ideas, and drafting temporary text.

Importantly, Vincke stressed that the shipping assets in the game will be crafted entirely by actual writers. "Larian is creating every line ourselves," he stated.

We are actively increasing our roster of concept artists and are actively forming narrative groups.

Given that visual development is being explicitly called out — we presently have 23 visual developers and have roles to fill for additional creatives.

Everything we do is additive and focused on having people spend more time on the creative process.

Any ML tool applied correctly is additive to a creative team process, never a stand-in for their craft.

Responding to Feedback and Defining the Path

The admission of AI usage initially sparked concern among portions of the player base. In response, Vincke issued more clarification on public forums.

"We use AI tools to gather inspiration, similar to we use search engines and physical media," he stated. "During the very early planning process we use it as a basic framework for structure which we then replace with hand-crafted concept art."

He added, "We've hired creatives for their unique talent, not for their willingness to replicate what a machine suggests."

Focused Uses for Machine Learning

Vincke had previously detailed the company's targeted approach to machine learning, defining its use into three main areas:

• Automation of Tedious Tasks: Areas like motion capture cleaning, voice editing, and Larian-specific work like adjusting assets for various species.
• Fast-Tracked Experimentation: Using tools to speedily create simple versions of mechanics to test concepts before full production.
• Experimental Frontiers: Investigating how machine learning could one day facilitate new forms of player agency, especially in creating dynamic reactions in a detailed game universe.

He clearly stated that core creative disciplines — like writing — are are absolutely not departments where the studio is cutting creative input. Conversely, Larian is recruiting more in these very roles.

"Larian is not shipping a game with machine-made assets, and we are certainly not considering trimming down staff to replace them with AI," Vincke concluded.