The Divinity Developer Explains Its Use of Machine Learning for Upcoming Divinity

The studio behind popular titles like Baldur's Gate 3 and Divinity: Original Sin just teased its next major project, generating a wave of excitement within the player base. However, subsequent statements from the studio's figurehead have added clarity to the conversation, focusing on the studio's approach toward machine learning.

A Tool for Ideation, Not Replacement

In a new clarification, Larian's director detailed that the developer is using machine learning for particular supporting purposes. These include fleshing out PowerPoint slides, producing early-stage concept art, and creating temporary copy.

Importantly, Vincke emphasized that the end content in the game will be authored solely by actual artists. "Larian is developing everything ourselves," he stated.

We are continuously growing our team of storytellers and are busily putting together writing teams.

As concept art is being explicitly mentioned — we presently have 23 visual developers and have positions available for additional creatives.

Each initiative we do is incremental and focused on letting our team spend additional energy on making content.

Any ML tool applied correctly is a boost to a artist's workflow, not a replacement for their craft.

Addressing Concerns and Clarifying the Vision

The admission of AI usage at first sparked backlash among some the fanbase. In reaction, Vincke offered additional detail on online platforms.

"We use AI tools to research ideas, similar to we use search engines and reference books," he stated. "During the conceptual brainstorming phase we use it as a rough outline for composition which we then swap out with hand-crafted concept art."

He added, "We've hired artists for their creative vision, not for their willingness to follow what a algorithm proposes."

Three Pillars of Practical Application

Vincke had in the past outlined the team's focused approach to this technology, categorizing its use into three main areas:

• Handling Monotonous Jobs: This includes polishing mocap data, dialogue cleanup, and technical processes like retargeting animations.
• Fast-Tracked Experimentation: Using systems to rapidly prototype rough mock-ups of gameplay ideas to validate concepts prior to expensive implementation.
• Experimental Frontiers: Investigating how machine learning could one day facilitate emergent gameplay, specifically in managing player-driven narratives in a complex RPG.

He specifically stated that key artistic domains — such as writing — are are absolutely not fields where the studio is reducing human involvement. On the contrary, Larian is expanding its staff in these precise fields.

"Larian is neither shipping a game with any AI components, nor looking at cutting teams to swap them out with AI," Vincke summarized.