AI Tools for Game Art Development
The landscape of game art production is rapidly evolving with the emergence of powerful artificial intelligence (AI) tools. As more and more AI tools emerge, we need to determine their usability and effectiveness in Game Art Production. With multiple facets in Game Art Production, such as Concept Art, 3D, 2D, Animation, UI, Marketing art, etc, how can the available tools help supplement these areas and speed up production? Let’s take a look!
2D AI TOOLS
One area where generative AI (GenAI) tools help out a lot, is in the concept stage. Using tools like Chat-GPT, RunDiffusion, and Midjourney, we are able to come up with quick ideas and concepts as to what style we want to achieve, what environment or characters would be cool to create, or what mood we want to portray. We can get a wide variety of ideas in a short amount of time.
Of course, we can go beyond using AI for concept art. Tools like Ideogram can assist in quickly iterating on different directions with logos and allow us to experiment with the different types of branding we want to go for. This does a lot for a game’s visual identity.
For marketing materials and key visuals, AI tools like ComfyUI optimize the creation process through highly adaptable workflows. These tools allow for precise control of image parameters — such as character count, support for switching poses, and overall scene composition — while maintaining or switching between multiple style guidelines. Such streamlined pipelines are invaluable for producing consistent yet versatile marketing assets like static ads, splash screens, and teaser images.
Adding on, we can also train our own AI models with Scenario. This tool enables us to create models tailored to our game’s unique art style. Imagine feeding the AI concept art for your characters and environments, and then using it to generate variations or even entirely new assets that look consistent with what we have. We’ve utilized this tool in creating renders and marketing material for Mighty Action Heroes.
As you can see from the pictures above, Scenario is a potent tool that can give usable results. The tricky part here is knowing exactly what we want to create and how we can best utilize the tools to create them. It does take some amount of time to get an understanding of what we want to generate and then establish a workflow to realize it. As much as AI tools can help us, the task still falls to the artists to help edit, tweak, package, and present something for it to be considered usable or complete. Finer edits, subtle layout changes, color correction, and overall composition of the final asset are tasks that AI is unable to do well just yet.
3D AI TOOLS
Now, in terms of 3D, can AI help us create 3D models from scratch? How well can it do that? Can it make variations? Can it replicate styles or be consistent? What's the amount of user control we have? What about texturing? Can it generate UVs? Can it texture the model for us? Is it accurate?
Let’s look at a few 3D AI tools available and see how well they tackle our problems. Note that these software are still being developed, so they may have improved or added new features after the time of writing.
A few 3D tools I’ve tested include Tripo, CSM, LUMA, and RODIN-1. These tools allow us to instantly generate models of varying quality via text or image input. Take a look at some of the models being generated below.
Generally, we can see that different software is slightly better at different things. All of them are pretty good at generating a rough mesh. Some are more accurate to the reference than others. It's hard to generate something with a usable topology. A lot of AI tools don't have that nuance or the technology to do so yet. More so if we want a character to be created a certain way (special topology/ separated joints/ separating clothes from the model). AI tools just aren’t capable of it as of now.
As we can see, 3D tools are currently not useable as-is in production. It doesn't have that polish of good topology, UVs, texture generation, etc. That said, there are some areas where 3D generative tools can be applied — rapid prototyping, idea generating, quick testing, smaller props, etc. Think of using it as a first pass. We can quickly use them as rough ideas of what we want to create or test how characters or objects would look in-game.
There are other promising aspects of 3D AI tools. Stable Projectorz and CSM have AI-assisted texturing tools which can help generate or enhance textures for your 3D models. It projects an image based on the model’s orientation from which we can select and paint on the parts of the textures we want to be applied.
Of course, there are still issues with it. The textures are only applied based on one view and the blending of textures can be improved. At the same time, this shows potential. As AI tools improve, we would be able to use them more to aid game art production.
Recently there’s been an interest in using hybrid workflows. This means using rough 3D models to help determine 2D generation’s form and structure better. We can use the 2D generation as a “filter” to give different looks. This type of workflow is still ver y very rough, but it does open the door to finding and using different workflows that combine 2D and 3D.
Closing Thoughts
While AI for game art production is still in its exciting infancy, with companies like Autodesk entering the fray with Generative 3D AI tools, the future holds great promise. Utilizing these AI tools ultimately boils down to each development team’s specific needs and project scope. However, the ability to generate rough meshes in a flash offers a valid option for prototyping, drafting, and rapid testing. This allows artists to jump right into iterating on ideas and refining execution.
Just like the established AI tools for auto-rigging, texture baking, and auto-UVs, utilizing AI as a collaborative tool across various production stages makes sense. While these tools streamline workflows and free up artists’ time, they cannot fully replace the human touch. The discerning eye for detail, the ability to fine-tune, and the overall vision for the desired result remain firmly in the artist’s domain. AI serves as a powerful assistant, not a replacement, on the ever-evolving canvas of game art creation.
