In the gaming world, personalization and character customization, supported by in-game microtransactions, have driven some of the most successful game titles. Take, for instance, “Player Unknown’s Battlegrounds” (PUBG), which garnered $744 million from weapon customizations alone. Players crave the ability to create caricature avatars that reflect their individuality. This trend is evident in a study on Miis in the Wii gaming system, where most individuals customize their avatars to resemble themselves.
However, traditional customization options often fall short, providing limited possibilities constrained by pre-made choices derived from market research. This is where Stable Diffusion has gained traction, offering a breath of fresh air with its potential for truly limitless customizations, including caricatures from photos.
What does this offer for businesses and gamers?
- Traditionally, businesses and gamers have been limited in the variety of props they could apply to their game avatars before starting a gameplay session. With generative AI, gamers have infinite possibilities to create and use new props in real-time, significantly enhancing their experience. They can customize their avatars with just a few prompts, creating unique styles effortlessly. This innovation can be monetized by charging players per generation or based on the time usage of the model to create more complex game artifacts.
- Gamers can trade or share these props, and if implemented correctly, these artifacts can be minted as part of a blockchain, such as in web3. Developing a platform for their exchange could further monetize this system through an “exchange fee.” Data Reply has experimented with such a platform using the Ethereum blockchain to transact artifacts in Minecraft, showcasing the immense potential generative AI offers, especially with its wide variety. Additionally, the best caricature artist can create unique digital art pieces, enhancing the value and uniqueness of these artifacts.
In this article, we will walk you through the utilities and technologies that can be leveraged to develop such a platform. We have experimented with a few avatars as part of our Generative AI R&D. But before diving into the development process, we need to understand the AI model used. Let us dive straight in.
Understanding Stable Diffusion and How to Get Started
Stable Diffusion, an advanced deep learning model by Stability AI released in August 2022, enables users to generate and modify detailed images with creativity. It employs a latent diffusion model and offers three main features for content generation: `txt2img` for generating images from text prompts, `img2img` for modifying images with text prompts, and `depth2img` for using text prompts and depth information to alter images.
The model also supports inpainting to fix or generate specific parts of images. Numerous parameters and open-source checkpoints are available to fine-tune and achieve diverse aesthetic styles.
Implications and Thoughts
Integrating generative AI into gaming offers various benefits, like enhanced avatar customization and more inventive virtual worlds. However, challenges include managing inappropriate content and current limitations in content generation. Despite these issues, rapid advancements in AI technology promise improvements, including higher-resolution image rendering for gaming.
Future Work and Enhancements
The challenge is transforming 2D avatar customizations into 3D objects. Two techniques are explored:
- Using warping and predefined masks in Blender to map 2D images onto 3D objects. This method may introduce complications due to lighting angles and requires a predefined Blender avatar.
- Utilizing Neural Radiance Fields (NeRF) to create 3D files from multiple angles of photographs. This approach helps with significant shape changes but faces complications due to the probabilistic nature of Stable Diffusion, although it can estimate depth from single images when combined with other models.
About The Author
