Installation Guide

Get up and running with Radiance v2.3.3 in minutes.

⊞ Windows Installation

For Windows users running standard ComfyUI portable or a Git clone.

1. Clone the Repository

Navigate to your ComfyUI custom nodes directory and clone:

cd ComfyUI\custom_nodes
git clone https://github.com/fxtdstudios/radiance.git

2. Install Dependencies

Use the embedded python in your ComfyUI portable environment:

cd radiance
..\..\..\python_embeded\python.exe -m pip install -r requirements.txt

If you are using a standard Python environment instead of portable, just use pip install -r requirements.txt.

3. Restart ComfyUI

Launch ComfyUI. Check the terminal for the [FXTDStudio Radiance] Loaded 79 nodes confirmation.

 macOS (Apple Silicon M1/M2/M3 & Intel)

Radiance supports Apple Silicon via PyTorch MPS (Metal Performance Shaders) natively. The GPU acceleration will automatically switch from CUDA to MPS if it detects a Mac.

1. Clone the Repository

Navigate to your custom nodes directory:

cd ComfyUI/custom_nodes
git clone https://github.com/fxtdstudios/radiance.git

2. Install Dependencies

For macOS, you may need to use pip3 or your conda environment's pip:

cd radiance
pip install -r requirements.txt

MPS Memory Note

For heavy Radiance nodes operating on 4K EXR files or using the `Flux Radiance Sampler`, ensure your Mac has at least 16GB of Unified Memory, as PyTorch on Metal can be memory intensive.

🐧 Linux Installation

For Linux users (Ubuntu, Arch, etc.) using Docker or local venv installations.

1. Clone the Repository

cd ComfyUI/custom_nodes
git clone https://github.com/fxtdstudios/radiance.git

2. Install Dependencies

cd radiance
pip install -r requirements.txt

Note: If you encounter issues compiling `colour-science` or `opencv-python`, make sure your system has the standard build-essentials installed (sudo apt install build-essential libgl1-mesa-glx).

◎ System Requirements

Item/Library	Status	Description
CUDA or MPS GPU	Required	Radiance performs heavy 32-bit math. A GPU with at least 8GB VRAM is highly recommended.
colour-science	Required	Advanced ACES, chromatic adaptation, and spectral calculations.
opencv-python-headless	Required	Used for certain fallback image operations and EXR writing when OpenEXR is missing.
OpenColorIO	Optional	Required for OCIO color transforms.
OpenEXR	Optional	Required for reading/writing native multi-channel .exr files.

◎ Nuke Bridge Setup

Firewall Warning

The bridge uses Port 5555 for streaming the EXR buffer. Ensure your OS firewall allows local traffic on this port between Python and Nuke.

Back to Home View Nodes