NVIDIA Unveils DLSS 5: Promising Photorealistic AI-Powered Graphics This Fall

Just months after its DLSS 4.5 announcement at CES, NVIDIA has revealed its latest leap in upscaling technology: DLSS 5. The company is significantly enhancing its AI integration for this new iteration, asserting that DLSS 5 will "infuse pixels with photoreal lighting and materials" using a real-time neural rendering model, slated for release this fall.

What practical implications does this advancement hold? During NVIDIA’s GTC 2026 keynote, CEO Jensen Huang demonstrated the technology's capabilities using game titles such as Resident Evil: Requiem, Hogwarts Legacy, and Starfield. DLSS 5 noticeably improves the detail in character hair and skin tones. However, it's worth noting that these comparisons were made against versions of the games running without any DLSS features enabled. The extent of its improvement compared to DLSS 4.5 with advanced features like path tracing remains to be fully clarified.

NVIDIA elaborated in a blog post, stating, “DLSS 5 processes a game’s color and motion vectors for each frame as input, leveraging an AI model to imbue the scene with photorealistic lighting and materials. These are firmly anchored to the source 3D content and maintain consistency across frames.” The company also confirmed that the technology operates in real-time and supports resolutions up to 4K.

Huang's demonstration of DLSS 5 utilized a system equipped with two RTX 5090 GPUs. While it will eventually be compatible with a single video card, it's anticipated that such a card would require performance comparable to dual 5090s. Huang also positioned DLSS 5 as a crucial step toward achieving Hollywood-caliber real-time rendering without the extensive GPU power typically demanded by professional studios. This innovation appears to resemble a generative AI video model that offers direct control to developers, moving beyond mere AI prompts.

In a characteristic display of self-praise, NVIDIA has hailed DLSS 5 as the "biggest breakthrough in computer graphics" since real-time ray tracing emerged in 2018. Yet, given that real-time ray tracing itself has yet to become universally adopted by most gamers, it will be interesting to observe the level of enthusiasm for NVIDIA's AI-generated pixels.