A developer known as MAAN has successfully ported NVIDIA DLSS 5 Neural Rendering to run within a web browser environment using WebGPU. The project demonstrates compatibility with macOS, and the creator has provided a live demo that allows users to test the technology with their own models.
The implementation remains unconventional as NVIDIA does not officially support DLSS via WebGL or WebGPU. Standard DLSS operation relies on the NGX interface or the Streamline SDK, which require direct access to DirectX or Vulkan pipelines, including motion vectors and depth data. The developer has not yet disclosed the specific technical methodology used.
Performance testing on an RTX 4090 indicates that loading the neural model takes approximately one to two seconds. While the technology functions on Apple Silicon hardware, the current execution speed is limited. The developer plans to release the source code on GitHub in the near future.
Although the current state of the project is not suitable for real-time gaming, it offers potential utility for 3D model previews where latency is less critical. This proof of concept marks the first instance of DLSS 5 operating outside of NVIDIA's officially supported software paths.
