A ComfyUI workflow for applying the unofficial DLSS 5 Neural Rendering effect to either a video or a still image.
Choose your mode: Set the select value on both Switch input and Switch output to the same number:
1 = Video: Upload a clip in Load Video (Upload). The video branch uses temporal mode and saves an MP4.
2 = Image: Upload a picture in Load Image. The image branch uses still images mode and saves an image.
Keep the two switches matched: 1 and 1, or 2 and 2. Adjust the settings in the corresponding DLSS 5 Neural Rendering (Unofficial) node, then queue the workflow.
Required custom nodes: DLSS 5 Neural Rendering (Unofficial), Impact Pack switches, Video Helper Suite, rgthree, and the additional preview/timer nodes shown in the workflow. Replace the example input files with your own before running.
DLL setup: You must obtain a compatible
nvngx_dlssnr.dllfrom a source you’re authorised to use, then place it atComfyUI/custom_nodes/ComfyUI-DLSS5-NR/runtime/nvngx_dlssnr.dll. Restart ComfyUI and check the DLSS 5 NR Runtime Info node. The ZIP includes the node’s other required DLLs; it does not include NVIDIA’s Neural Rendering DLL. GitHub
The project maintainer does not provide a download link for nvngx_dlssnr.dll, so I wouldn’t put a random DLL mirror in your Civitai pos.
play with node 1 for skin and lightness
Description
Initial release. Includes image and video branches; set both switches to 1 for video or both to 2 for image.
Comments (2)
Images look darker
yes you need to play with the node 1 to get a sweet spot