wilsjo2/OptiScaler-DLSSNR-PreSR-Multipass

A game mod that uses NVIDIA AI to change lighting, detail and colour, with controls for strength and performance.

What it solves

It provides a way to modify game visuals—specifically lighting, detail, and color—using NVIDIA AI, allowing users to customize the look of a game while balancing performance costs.

How it works

The mod uses NVIDIA's neural rendering technology (via the nvngx_dlssnr.dll model) to apply visual effects. It can be integrated into the game's rendering pipeline either before or after upscaling, and it supports multiple passes with different settings to refine the image. It also includes a noise reduction (NR) feature to mitigate green noise in DirectX 12 games.

Who it’s for

PC gamers with NVIDIA RTX 20, 30, 40, or 50-series GPUs who want to enhance or customize the game's lighting and colors through AI-driven rendering.

Highlights

  • Customizable Visuals: Adjust strength, lighting, detail, and color settings.
  • Targeted Effects: Separate settings for skin and scenery.
  • Flexible Pipeline: Ability to run the effect before or after upscaling.
  • Multi-pass Rendering: Apply the effect multiple times for different results.
  • Model Resolution Control: Lower the model resolution to reduce performance impact.
  • HDR and Frame Generation Support: Works with HDR and frame generation in native DirectX 12 games.

Related

  • Project
    rgeirhos/Stylized-ImageNet

    A tool for creating Stylized-ImageNet, a version of ImageNet that distorts textures to encourage CNNs to prioritize object shape over texture for better robustness.

  • Project
    Visionary-Laboratory/visionary

    A WebGPU-powered platform for real-time rendering of Gaussian Splatting variants and 3D meshes directly in the browser using ONNX Runtime.

  • Project
    IvanDrokin/torch-conv-kan

    TorchConv‑KAN is a PyTorch library that implements convolutional layers where each kernel is a set of learnable univariate functions (KAN, Fast‑KAN, Cheby‑KAN, etc.). It ships with many layer variants, CNN‑style model families (ResKANet, DenseKANet, VGG‑KAN, U‑Net‑KAN), pre‑trained Imagenet‑1k checkpoints, and training scripts that use Accelerate, Hydra, WandB, and Ray‑Tune. The project is under active development and accompanies a research paper on Kolmogorov‑Arnold convolutions.

  • Project
    LSH9832/edgeyolo

    EdgeYOLO is an anchor-free object detector optimized for real-time performance on embedded edge devices like Nvidia Jetson, supporting multiple deployment formats including TensorRT and RKNN.