NVIDIA DLSS 5 Mod Runs In A Browser and Even Works on Apple Silicon

Web Demo: Capabilities and Constraints

A developer known as MAAN has developed a web implementation that functions as a model viewer capable of applying the DLSS 5 neural filter. Unlike the official release, this demo does not require a specific GeForce RTX 50 series GPU, as it leverages WebGPU to run on almost any hardware with the necessary support. This includes systems running Apple Silicon.

Practical testing shows the demo is quite versatile. It successfully ran on an ASUS ROG Flow z13 tablet equipped with Radeon 8060S graphics, and the developer confirmed compatibility with macOS. However, it failed to load on a Snapdragon 888-powered smartphone, likely due to memory limitations. The neural network itself requires a download of nearly 150 MB, which is substantial for a web application and may pose challenges for devices with limited resources.

Technical Implementation: A Clever Workaround?

It is important to note that this is not a real-time rendering engine, but a viewer that applies the DLSS 5 neural filter to pre-loaded models. Since NVIDIA does not offer official support for WebGL or WebGPU, the implementation is technically remarkable. According to Whycry, the developer likely either extracted the neural network weights to reimplement them or created a shim to use the leaked binaries directly.

This experiment highlights the flexible nature of DLSS 5. Because it functions primarily as an AI-powered post-processing filter, it can be adapted to run on various hardware architectures. This follows a trend where modders have successfully implemented DLSS algorithms on older GeForce cards, as well as on AMD and Intel integrated graphics, proving the technology's potential utility beyond its official hardware requirements.