GMAN
RenderMan is the interface Pixar published for turning a 3D scene description into an image: cameras, geometry, lights and shaders, all in a plain-text RIB file. It's the API behind decades of film rendering, but the implementations that speak it are proprietary, heavyweight, or both.
GMAN is a from-scratch, RenderMan-compatible renderer. Point it at a .rib file and it produces an image using the familiar object → world → camera → screen → NDC → raster pipeline.
It implements analytic normals, backface culling, z-buffering, lighting (ambientlight, distantlight, pointlight) and C++ surface shaders including matte, plastic and metal. Output can be written as TIFF, PNG, JPEG or PNM.
The codebase is deliberately small enough to read end to end. It builds with CMake and a C++23 compiler, and is intended as a practical way to explore how a REYES-era rendering pipeline turns geometry into pixels.
Build and install
Needs CMake 3.25 or newer, a C++23 compiler (clang 16+ or gcc 13+), libtiff, libpng and zlib. POSIX only: macOS and Linux. A VS Code devcontainer is included.
git clone https://github.com/jac18281828/gman.git
cd gman
cmake --preset dev
cmake --build build --parallel
Render your first scene
A RIB file is plain text. Save this as sphere.rib:
Display "sphere.tif" "file" "rgba"
Format 640 480 1
Projection "perspective" "fov" [45]
Translate 0 0 5
WorldBegin
Sphere 1 -1 1 360
WorldEnd
Then render it:
./build/gman sphere.rib
sphere.tif lands next to it.
Where to find it
History
GMAN started in 1999, was shelved in 2002, and came back in 2026 on a CMake / C++23 build. The repository is public, licensed LGPL-2.1-or-later.