Nvidia unveiled DLSS 5 at GTC , a groundbreaking technology that uses generative AI to transform video game graphics in real time. CEO Jensen Huang called the development a “ GPT moment for graphics ,” combining traditional rendering with artificial intelligence to dramatically improve visual realism. This innovation represents a radical shift in how graphics are processed, shifting the emphasis from simply increasing resolution to creatively recomposing visual elements.
See also: Starfield: Adds Nvidia's DLSS in its upcoming patch
Unlike previous versions of DLSS that used machine learning for upscaling, DLSS 5 applies generative AI to completely recreate the lighting and materials in games. Examples from titles like Resident Evil Requiem , Starfield , Hogwarts Legacy , and EA Sports FC show impressive improvements in realism, but have also drawn criticism for distorting the creators' artistic intent. The technology represents a fundamental shift from the traditional super-sampling approach to a more creative and intrusive method of improving graphics.
According to Nvidia , the AI model is trained to understand complex scene elements such as characters, hair, fabrics, and transparent skin, as well as environmental lighting conditions by analyzing a single frame. DLSS 5 uses this understanding to create visually accurate images that handle complex elements such as subsurface scattering in skin and light-material interactions in hair. The model is trained on Nvidia ’s AI supercomputers , using massive databases of game scenes to ensure accuracy in reproducing different types of materials and lighting conditions.
See also: Nvidia CEO denies he's 'unhappy' with OpenAI – Big investment coming
Controversial changes to characters with DLSS 5
The most obvious effect of DLSS 5 is the significant changes to the appearance of the characters. In Resident Evil Requiem, protagonist Grace Ashcroft looks completely different, as if she has applied an Instagram filter that makes her lips fuller and adds intense shadow to her eyes. Similar effects are seen in Starfield, where characters appear overly lit despite being in environments that do not warrant such lighting. These changes have sparked heated debate in the gaming community, with many expressing concerns about the authenticity of the creators' artistic vision.
Todd Howard, head of Bethesda Game Studios, expressed his excitement, stating: “When NVIDIA showed us DLSS 5 and we ran it in Starfield, it was amazing how it brought it to life.” However, not all developers agree. Mike Bithell called the technology “nonsense” that removes any artistic direction from the gaming experience. This dichotomy reflects a broader issue in the gaming industry about how AI should be used to enhance rather than replace human creativity.
DLSS 5 can run in real-time at resolutions up to 4K , using the fifth-generation Tensor Cores of RTX 50 Series GPUs . The technology promises up to an 8X FPS multiplier by creating three AI-generated frames for each rendered frame, while reducing VRAM usage . This performance is achieved thanks to the integration of advanced transformer models that have been specifically trained to understand and reproduce complex visual scenes.
See also: Nvidia displaces Apple from TSMC's production capacity
Technical specifications and compatibility of DLSS 5
To fully utilize DLSS 5, gamers will need RTX 50 Series or higher, as the technology relies on Tensor Cores for optimal performance. Nvidia has integrated the technology into plugins for Unreal Engine 5 and Unity, making it easy for developers to adopt it. Users can adjust DLSS 5 through GeForce Experience or in-game menus, choosing between different levels of quality and performance depending on their preferences and hardware capabilities.
