The Mighty Bear Stance Towards the Use of GenAI

The general lack of ethical guidelines and regulations in the GenAI space has been a controversial topic that we were deeply concerned about and discussed intensively at Mighty Bear Games. We believe that while GenAI has a place in the future of game development, we must be firm with the rules of how we use it:

• We will use the technology as a tool to benefit the team, not replace it.

• We will not infringe copyrights and intellectual property, and will only use data that belong to us to train models.

Our strong belief in practising GenAI ethically has led us to be transparent about how we use the technology. We have also been actively collaborating with external partners to train models with existing art we created, to generate new content to ensure the generated artwork is still wholly made by Mighty Bear artists.

What “1 Character” in Mighty Action Heroes looks like from an art production perspective

In the context of Mighty Action Heroes, “1 character” means:

• 1x High-poly, marketing-quality character,

• 1x In-game “Hero Lounge” quality character,

• 1x In-game battle quality character,

• 1x NFT icon,

• 1x In-game portrait icon,

• 8x Marketing art renders

And that’s not including pre-production content like concept turnarounds and additional materials required for 3D production. What this means is that “1 character” is a lot of time spent and assets created for not-as-much consumable content for players.

The 3D Character Art Production Pipeline, Pre- and Post-AI Evolution

To break all that information down into simpler terms, it would take eight weeks across all six art disciplines to create just one character for the game, from concept to implementation and marketing. Considering the number of characters currently in Mighty Action Heroes, it would have taken us 128 man-weeks (almost 2 1/2 years!) to get all the work done.

That’s a lot of time!

Of course, we were able to get the bulk of the high-poly 3D models created concurrently with the help of our outsourcing partners, but the team still had to run a number of mundane, time-consuming but absolutely necessary technical processes known internally as the Hero Generation Pipeline to create, prepare and implement in-game assets, in-house.

Steps in the Hero Generation Pipeline to get final in-game character assets created. Ideation, concept and high-poly 3D model creation stages are not included in this diagram.

That’s a lot of repetitive work for a team with one to two artists in each discipline, but we also knew that AI could help to reduce that amount of work, end-to-end. Here’s how we did it:

Staying focused on creativity

The Concept Art Team was able to accelerate the creative process by using GenAI to assist in the ideation and pre-viz phases, helping us to stay more focused and hone in on our ideas, getting them approved by stakeholders more effectively.

Eliminating repetitive, mundane processes

With a growing Hero roster for Mighty Action Heroes, we turned our gaze towards building our own custom model to help reduce the bulk of repetitive processes in 3D production. To make up for the lack of technical expertise, the 3D team has been hard at work with testing and giving feedback to external partners for AI-driven tools that help with custom mesh generation and LoD generation for in-game assets. As things stand, I’m excited to share that we just started real-world use case testing with positive outcomes close to our target time reduction, and are looking to push for even more time saved.

LoD Generation is a simple and straightforward task, but it’s a time-consuming effort that doesn’t require any artistic creativity.

Reducing the number of clicks to achieve similar outcomes

We’ve also made leaps and bounds in in-house art tool development since Quinnie documented her quest for efficiency, and along with Technical Artist, Wayne, they successfully contributed to a 300% increase in productivity thanks to the tools developed for Version 2 of the Hero Generator Pipeline. These tools include features to automate mundane processes like skin-weighting, texturing, and exporting. With the amount of time already saved in our 3D character production pipeline, we’ve already started discussions on action points for Version 3 and should be able to start tool development sooner rather than later to further streamline our processes.

Casting our net far and wide to explore other frontiers

On less static fronts, we are looking into AI-powered video-to-animation solutions to help us keep up and stay relevant with social media trends. While the research objective here is not specifically for game art production per se, Quinnie has been doubling up as our marketing art support specialist on top of her main responsibilities as our resident Bear-nimator as the Hero Generator Pipeline also caters to produce content for our marketing collateral.

Chang Long, the Dragon Master: Early concepts (top) and final production assets (bottom from left to right) — LOD 1, LOD 0, marketing-quality model, NFT Icon and in-game hero portrait

“What do we do with the time saved?”, you might ask. At Mighty Bear Games, we are big on learning and growth, so we have spent the time freed up through the use of GenAI or tools created through AI to upskill the team and explore other avenues of creativity, such as Unreal Editor for Fortnite.

Conclusion

As an emerging tech, there are still many unknowns in the realm of AI, but this also means the world is our oyster.

Through the introduction of AI to our art team, we have successfully lowered skill barriers by having artists use ChatGPT to create scripts that help to make art production more efficient. In turn, our newly proficient artists use AI to create tools to streamline our art production pipeline, resulting in the team doing a lot less of the repetitive stuff, and having more time available to focus on creative works and upskilling more efficiently.

However, the technology is still in its infancy and most of the art we’ve generated through our R&D efforts is unusable for game development. We’ve used AI tools across our pipeline, but until we can get consistently good results, the human mind and hands are still king.

We often forget that Artificial Intelligence is exactly that, an “Intelligence” incapable of originality and creativity that draws from pre-existing, human-made content. GenAI is a powerful tool to help with the heavy lifting, eliminating time-intensive, mundane processes in art production pipelines, but human input is still necessary to create original content that excites our audiences.

How have you used AI as an artist, and how has it changed your processes? What’s your stance on using AI for game art production? Feel free to share them so we can all work towards a healthier, more ethical AI-powered future!