NVIDIA has announced DLSS 5, the next evolution of its Deep Learning Super Sampling technology, claiming it can achieve photorealism through a new real-time neural rendering model. Unlike previous DLSS versions focused on upscaling and frame generation, DLSS 5 uses AI neural inference to generate a significant portion of the final image, including lighting reconstruction and material rendering. Digital Foundry notes impressive handling of materials like metals, cloth, and foliage lighting, but flags it as still a work-in-progress with some screen-space errors. A key controversy surrounds character face rendering — early demos in games like Resident Evil: Requiem and EA Sports FC show dramatically altered faces, raising concerns that DLSS 5 changes artistic intent rather than simply enhancing fidelity. NVIDIA says the technology is semantics-aware and distinct from generative AI, and has developer support from studios like CAPCOM. A full launch is expected later in 2026 after three years of development.
Sort: