Larian Studios and Generative AI: How AI Is Shaping the Gaming Future

Honestly, I will be frank, I did have such gut reaction when I heard that Larian Studios is leaning into generative AI. Another studio as a substitute of artists with algorithms? However, when I took the trouble to find out what they are doing, it is quite the opposite.

This is what I discovered after taking unreasonable time to research generative AI implementation in larian studios.

They’re Not Making AI-Generated Games

To begin with, their real games do not have any AI content. Zero. Every concept art in Baldur’s Gate 3? Produced by their 23 human concept artists. Every line of dialogue? Written by humans. Voice acting? Professional actors.

CEO Swen Vincke was very explicit about this: “We are not launching a game with AI elements in it and we are not considering staff reductions to replace humans with AI.

So what is AI being used by them?

The Stuff Nobody Wants to Do

It turns out that Larian employs AI as the uncritical high flat rate work that consumes development time. Motion capture cleanup, voice editing post-processing, animation retargeting – in short, all those things that Vincke points out that no one wants to be involved with.

This is an interesting part: animation retargeting had previously required 2-3 days per character variant. Now? About an hour. It would save a ton of time particularly with RPGs where you can have dozens of characters which come with dozens of species.

Just to make a walking animation to fit a new type of body, an animator takes three days to get it right by hand, whereas AI takes about 60 minutes. The animator is able to compensate that two days in the process of making something creative.

White Boxing and Rapid Prototyping

The other big use? And there is the so-called white boxing, or that first step when game developers are very fast at visualizing the concepts of a game before investing any resources in them.

AI assists them in testing various options of layouts, characterization, and mechanics of the game much quicker. What would take days to occur is now done in hours. It is not taking over the creative vision, just accelerating the does that even work. phase.

Also they apply AI as a placeholder text when developing and as a reference tool when considering an idea – to a point of a supercharged Google image search. Once again the finishing touch is done by humans.

They are in fact employing more artists.

The next best thing is that, during the usage of the AI tools Larian has 530+ workers and continues to recruit. It has a total of 72 artists in their art department. They are creating writer rooms, recruiting actors, recording them, and employing translators.

The opposite of that is what happens when AI is employed by the studios to reduce expenses.

The Troubles They are Undergoing.

Everything does not sail smoothly. Their AI narrative validation system, which determines whether all the variations of the Baldur, Gate 3 story makes sense, is not a perfect one yet. It’s a work in progress.

And Vincke acknowledges that AI has not really significantly enhanced efficiency in general. The actual advantage is not needed to accelerate the development; freeing the creative individuals of mundane chores.

Other issues in the industry are larger, such as copyright issues regarding the AI training data, and how to comply with the GDPR concerning player data and determining on who should own AI generated work.

How You Can Use This Approach

Possibly the most accurate way to understand larian studios generative ai algorithms is that at minimum one of the following works:

Start with the tedious stuff. Identify activities that consume time and do not require innovativeness. Animation retargeting is done on DeepMotion. Mocap data can be cleaned up using various tools or you can deal with the generation of placeholders.

Use free resources to learn. There exists a free prompt engineering tutorial of W3Schools. On YouTube there are demonstrations of implementation. Reddit and Discord exchange real-life experiences.

Have human beings handle creativity. There are tools such as Ludo.ai or Scenario.gg that can be useful in prototyping and asset generation, but they should be treated as a starting point. The ultimate creative choices? That’s all you.

Reinstate the option, not obligatory, of AI. Larian gives the developers the choice of using these tools. It makes them less difficult and allows individuals to digest them at their own speed.

What’s Next for Larian

Another element of his vision is that 3-4 year development cycles can be achieved by deploying AI rather than 6 or more. Nevertheless, here is what he makes out to be his most important point: What teams like us will make the difference is the fact that we will continue to still have a desire to make something special on top of that.

The creativity of human beings remains more at the center. AI only has to be a cop to all the dull stuff.

The Real Takeaway

Having ransacked all this, the important part is the following: Larian does not focus on people substituting generative AI or reducing corners. It is about allowing innovation in the hands of creative employees.

Animation retargeting takes days down to an hour? Fine, now animators have more time to polish more animations or experiment with new ideas. Prototyping happens faster? Huh, that is cool, designers now have the opportunity to test more ideas and learn to discover the good ones much faster.

The difference in AI in the process of development and AI in the final game content is important. One frees up human talent. The other tries to replace it.

It is on one side of that line that Larian discovered is sensible. And honestly? It is likely that this is why Baldur’s Gate 3 has become what it became.

Read:

How to Level Up Fast in WoW Classic!