Update October 12, 2024: Acknowledging the mixed reception of my previous piece, I appreciate the feedback and recognize the concerns surrounding generative AI’s impact on the creative industry. While my enthusiasm for advancements in graphics and game technologies is genuine, I must emphasize the real consequences of AI on livelihoods in the creative sector. My aim was to explore the potential of AI in enhancing graphics through technologies like Freestyle, Frame Gen, and DLSS, but I understand the need for a balanced perspective.
Original story October 11, 2024:
The vision of photorealistic gaming is closer than ever, yet it may not emerge from the hands of traditional game developers or the innovative minds behind future game engines. Instead, the evolution of gaming graphics appears poised to be driven by artificial intelligence.
Before dismissing this notion due to the growing fatigue surrounding AI-generated content, consider the exciting possibilities it presents for creating games that closely resemble reality. YouTuber Soundtrick has been experimenting with existing game footage, utilizing the Gen-3 video-to-video tool on Runway ML. This process generates new video content based on the input footage, allowing for a reimagined take on various games. Although the results can be somewhat rough around the edges, the potential for generative AI to enhance graphics is evident, especially with sufficient processing power and real-time capabilities.
As an avid football fan and player of FIFA/FC, I often ponder why the graphics don’t mirror the real thing. With AI trained on countless televised football matches, there exists a wealth of data to develop a model capable of rendering a game that looks indistinguishable from an actual match. This shift could redefine rendering techniques, moving away from traditional ray tracing in favor of AI-generated visuals. If the outcome is a more realistic representation, many would welcome this change.
In this new paradigm, the GPU’s role may shift dramatically. Instead of focusing on high-fidelity rendering, the graphics card might primarily handle low-resolution polygons and basic textures for character recognition and animation. The emphasis would then be on memory and matrix processing capabilities, essential for executing the complex AI algorithms necessary for generating photorealistic frames at rapid intervals.
While achieving low-latency responses in a game environment poses challenges, the prospect of such a reality would transform the gaming landscape. Currently, our gaming experiences are uniform, dictated by pre-baked models and textures crafted by developers. However, with the advent of generative AI, the visual experience could vary significantly based on individual preferences and hardware configurations.
Imagine a future where the next generation of DLSS employs a generative AI post-processing filter, resulting in distinct visual outcomes depending on the model and training data used. Even with identical GPUs, players could enjoy unique visual experiences tailored to their tastes. Nvidia’s existing Freestyle filters hint at this potential, enabling players to customize their gaming visuals further. The possibilities are tantalizing, especially when considering the potential to breathe new life into classic titles.
Wanna replay GTA V, but have a hankering for more of a Sleeping Dogs vibe? Tell the filter that you want it to be set in Hong Kong instead of Los Santos.
In this imagined future, layering AI post-processing over older games could yield stunning remasters or entirely new visual interpretations. Fancy a replay of GTA V with a twist? Perhaps you’d prefer a setting reminiscent of Sleeping Dogs, or even whimsical alterations like transforming characters into the faces of beloved childhood icons. While this evolution may disrupt the roles of modders and game artists, it also empowers players to shape their gaming experiences.
Despite the hurdles ahead—such as the need for substantial local storage and memory to support rapid processing—this future is not as distant as it seems. The prospect of revisiting classics like Half-Life with enhanced visuals is an enticing one, even if it comes with its share of uncanny moments along the way.