With the Game Developers Conference (GDC) officially kicking off in San Francisco, NVIDIA once again demonstrated its strong leadership in AI content creation. To help game developers and creators build cinematic worlds and characters more smoothly on NVIDIA RTX AI PCs and DGX Spark desktop supercomputers, NVIDIA today announced a series of new updates.
The highlights of this project focus on significantly simplifying the user interface of the most popular open-source generative AI tool, ComfyUI, and introducing RTX Video Super Resolution real-time 4K upscaling technology. It also releases NVFP4 and FP8 format support for the latest models from Lightricks and Black Forest Labs, resulting in explosive growth in both the performance and ease of use of local AI video generation.
Say goodbye to complex nodes: ComfyUI's brand new "App View" interface
In the past, although ComfyUI had a high degree of freedom and a strong open-source ecosystem, its complex "node graphs" interface, which resembled a spider web, often deterred many beginners.
To address this pain point, ComfyUI has launched a brand-new "App View" mode. This simplified interface hides the complex underlying logic, allowing users to quickly generate content simply by entering prompts, adjusting a few basic parameters, and pressing the generate button.
Of course, advanced creators will still retain the original node-based full operating experience (Node View). Users can seamlessly switch between the two modes, perfectly balancing the needs of beginners and professional customization.
Embrace NVFP4 and FP8 formats: Performance increases by 2.5 times, and display memory usage is reduced by 60%.
App View is perfectly compatible with ComfyUI's underlying optimizations for RTX graphics cards. Since last September, RTX GPUs have seen a 40% performance improvement on ComfyUI, and now ComfyUI goes a step further by natively supporting NVFP4 and FP8 data formats.
NVIDIA announced that the LTX-2.3 from Lightricks, as well as the FLUX.2 Klein 4B and FLUX.2 Klein 9B models from Black Forest Labs, are now available in NVFP4 and FP8 formats, respectively.
• Compatible with the NVFP4 format of the GeForce RTX 50 series:Video generation performance will increase by up to 2.5 times, while memory usage will be reduced by 60%.
• Use FP8 format:Performance is improved by 1.7 times, while memory usage is reduced by 40%.
Creators can now immediately enjoy this hardware-level performance bonus simply by downloading these new format checkpoints from Hugging Face and loading the preset workflow in ComfyUI's template browser.
30x 4K upscaling experience: RTX Video Super Resolution supports ComfyUI
In AI video generation, creators often face the impossible triangle of "speed, display memory, and control." Most creators tend to quickly generate low-resolution preview videos first, and then perform high-resolution upscaling after confirming that everything is correct; however, existing upscaling tools often take several minutes to process a 10-second 4K resolution video.
NVIDIA has now officially introduced RTX Video Super Resolution as a standalone node in ComfyUI. This technology directly utilizes the Tensor Cores of the RTX GPU for computation, processing data up to 30 times faster than popular local upscaling tools on the market. It instantly upscales generated videos to 4K resolution with minimal usage of display memory. Furthermore, NVIDIA has simultaneously released a free Python wrapper (Python Wheel) and sample scripts in the PyPI repository for AI developers, making it easy for them to integrate this powerful upscaling technology into their projects.
Bonus Feature: Other RTX AI PC Updates from NVIDIA at GDC 2026
In addition to the major update to ComfyUI, NVIDIA also shared several advancements in the RTX ecosystem during GDC 2026:
• LM Link operates across devices:It allows users to run the AI Agent on their laptops, but the actual inference computing power calls on remote DGX Spark or RTX desktops to execute LM Studio models, enabling a seamless local AI experience across devices.
• DLSS 4.5 officially launched:The NVIDIA App Beta update on March 31 will release "Dynamic Multi Frame Generation" and "Multi Frame Generation 6x Mode" specifically for GeForce RTX 50 series gamers and developers.
• RTX Remix particle effects upgraded:April will bring an update to the Advanced Particle VFX effects, allowing mod developers to create richer and more immersive particle effects.
• Topaz Labs NeuroStream Optimization:This proprietary display memory optimization technology enables highly complex AI models to run smoothly on consumer-grade NVIDIA graphics cards.
• VoiceMod supports Windows ML:It became one of the first applications to enable GPU inference using Windows ML, significantly improving the performance and sound quality of voice processing.
