A Blender render engine powered by MoonRay, DreamWorks Animation's open source production renderer. It works like Cycles: interactive rendering in the viewport (Rendered shading mode) and final renders with F12.
MoonRay doesn't support Windows, so it runs inside WSL2. Blender talks to it
through moonray_bridge, a small C++ server:
Blender (Windows) WSL2 (AlmaLinux 9)
┌───────────────────────┐ TCP localhost ┌───────────────────────────┐
│ moonray_blender add-on│ ───────────────▶ │ moonray_bridge │
│ · exports the scene │ RDLA + meshes │ one process per session │
│ · draws the frames │ ◀─────────────── │ MoonRay RenderContext │
└───────────────────────┘ RGBA pixels └───────────────────────────┘
Status: early (0.1.0), tested on Windows 11 + Blender 5.1. Expect rough edges.
- Interactive viewport with progressive refinement: orbiting the view, moving objects or editing materials updates the render live. Only what changed is sent (moving an object sends a matrix, not its mesh).
- Final render (F12) with progress in the Image Editor.
- Geometry: meshes, curves, surfaces, metaballs and text (evaluated with modifiers), with UVs, per-corner normals and one material per slot.
- Instancing: collection instances, particles and Geometry Nodes instances via
RdlInstancerGeometry. - Lights: Point →
SphereLight, Spot →SpotLight, Sun →DistantLight, Area →RectLight/DiskLight, World →EnvLight(color or HDRI). Intensities are converted so scenes match Cycles (verified with single-light test scenes). - Materials: Principled BSDF, Diffuse, Glossy, Glass and Emission are converted to
DwaBaseMaterial. Image Texture nodes linked directly to an input becomeImageMap, and the Normal Map node becomesImageNormalMap. - Cameras: perspective and orthographic, with shift, sensor fit and depth of field.
- Only textures linked directly to shader inputs are translated; intermediate nodes (Mapping, ColorRamp, Mix...) are ignored. No subsurface or sheen yet.
- HDRI orientation and instancing haven't been validated against Cycles yet.
- GPU: the add-on has a Device option (CPU / GPU) that selects MoonRay's XPU mode, but NVIDIA doesn't ship OptiX in the WSL2 driver (NVIDIA forum), so under WSL MoonRay falls back to CPU automatically. GPU rendering works on a native Linux install (see below).
Requirements: Windows 10/11 with WSL2, Blender 4.2 or newer, ~30 GB of free disk space and a few hours for the first MoonRay build.
wsl --install -d AlmaLinux-9 --no-launch
wsl -d AlmaLinux-9 -u root -- bash /mnt/c/<path-to-repo>/wsl/setup_wsl.sh
wsl -d AlmaLinux-9 -u root -- bash /mnt/c/<path-to-repo>/wsl/build_moonray.sh allAlmaLinux 9 is binary compatible with Rocky Linux 9, the platform MoonRay officially
supports. Build logs go to /opt/MoonRay/logs/.
wsl -d AlmaLinux-9 -u root -- bash /mnt/c/<path-to-repo>/wsl/build_bridge.shIt is installed to /opt/MoonRay/bridge/.
blender --command extension build --source-dir addon/moonray_blender --output-dir dist
blender --command extension install-file -r user_default --enable dist/moonray_blender-0.1.0.zip(or Edit → Preferences → Get Extensions → Install from Disk). Then pick MoonRay as the render engine in Render Properties.
The add-on starts moonray_bridge in WSL automatically the first time you render.
The WSL distribution, user and port can be changed in the add-on preferences.
wsl -d AlmaLinux-9 -u root -- bash /mnt/c/<path-to-repo>/wsl/install_cuda.sh
$env:WSLENV="GPU/u"; $env:GPU="1"
wsl -d AlmaLinux-9 -u root -- bash /mnt/c/<path-to-repo>/wsl/build_moonray.sh moonraywsl/patch_moonray_wsl.py (applied automatically by build_moonray.sh) lets MoonRay
detect the NVIDIA driver under WSL, but OptiX itself is still missing from the WSL
driver, as explained above.
blender --background --factory-startup --python tests/test_export.py # exporter, no MoonRay needed
blender --background --factory-startup --python tests/test_client.py # protocol, fake bridge
blender --background --factory-startup --python tests/test_render.py # real F12 render
blender --factory-startup --python tests/test_viewport.py # viewport, opens a window| Folder | Contents |
|---|---|
addon/moonray_blender/ |
Blender add-on (extension for 4.2+) |
bridge/ |
moonray_bridge (C++, built against MoonRay) |
wsl/ |
Scripts to set up WSL and build MoonRay and the bridge |
tests/ |
Tests that run inside Blender |
Messages have a 16-byte header (magic, type, size) followed by the payload.
Message types live in bridge/src/Protocol.h and
addon/moonray_blender/protocol.py. Small scene
changes travel as RDLA text; meshes travel as binary arrays.
- The Blender add-on (
addon/,tests/) is licensed under GPL-3.0-or-later, as required for Blender add-ons. moonray_bridgeand the build scripts (bridge/,wsl/) are licensed under Apache-2.0, the same license as MoonRay.
MoonRay is a trademark of DreamWorks Animation. This project is not affiliated with or endorsed by DreamWorks Animation.