Python: SPG Grayscale#

The “hello world” of Sensor Processing Graphs. A custom CUDA kernel converts the renderer’s LdrColor AOV to grayscale entirely on the GPU and publishes the result as a new LdrGrayscale AOV. This is the minimal three-file SPG shader (.cu kernel, .cu.lua launch script, .usda shader definition) wired into a RenderProduct.

“Create the smallest useful SPG example: a CUDA kernel that converts the LdrColor render output to grayscale, wired into a RenderProduct and read back from Python as a new AOV.”

SPG grayscale example output

Prerequisites#

  • Python 3.10-3.13

  • uv

  • An NVIDIA RTX-capable GPU and a supported driver

Running#

uv run main.py                                     # CUDA
uv run main.py --scene grayscale_scene_slang.usda  # Slang

The first step compiles the kernel and can block for up to 5 minutes on a fresh shader cache. A successful run writes _output/input.png and _output/grayscale.png, which is the same image with the colour taken out.

See the SPG reference for every Lua function and USD attribute you can author.

On this page