Roblox Reality is an AI-based video supersampler like DLSS 5

AI Video & Visuals


It appears that Roblox Corporation is currently working on an AI-based supersampler similar to DLSS 5 for Roblox called Roblox Reality. Roblox Reality aims to make your creations on Roblox look and feel more realistic.

According to the developers, Roblox Reality is a hybrid architecture that combines the power of the Roblox cloud and game engine with the photorealism of the video world model.

“Core world state is persistently and efficiently stored on servers, using cost- and space-efficient storage to ensure consistency across clients and support consistency across hours, sessions, and days. Multiplayer gameplay is supported by strong server privileges for fairness and consistency, and speculative client-side simulation for low latency. Rendering uses cloud-based Level of Detail (LOD) The Roblox video model (Super Upsampler) leverages rendered video and rich data model context to produce probabilistic visuals and impressive realism, running on the edge of any player with optimal performance from cloud edge GPU infrastructure.”

To introduce Roblox Reality, the developer shared the following video. The video in the top left is currently Roblox content recorded using the Roblox rendering engine. The video in the top right corner represents 3D data that can be used to adjust video generation. Next, the bottom left video shows the current Roblox upsampled video model running in the lab, but not yet running in real-time. Finally, the video on the bottom right shows a mockup of the product vision and what will be possible with this technology in the future.

Overall, Roblox Reality will be a hybrid architecture that splits duties between the Roblox game engine and the Roblox video world model.

The Roblox game engine handles the structural and logical aspects of the world, providing stable long-term memory, symbolic logic, and repeatable simulations. It is also responsible for basic physics operations such as collisions and movement. The main movement of objects is managed by the engine. For example, the car’s position and speed, wheels, shocks, and steering are all handled by the game engine. The video world model then overlays additional visual and generative components, such as beads of water flowing along the windshield or flapping leaves as the car zooms in, to provide breathtaking visuals.

In other words, it works exactly like DLSS 5. Roblox Video World Model is an AI-based post-processing that can be applied to the final image. Try to make the final image more realistic. Now, whether you like this approach or not is entirely up to you.

At this time, there is no timeline for when Roblox Reality will be available. Still, this shows that NVIDIA was once again ahead of the curve when it announced DLSS 5. So it will be interesting to see the reactions of all Roblox fans.

Look forward to future updates!



Source link