A Nuke Plugin for high-quality denoising of CG renders using Intel Open Image Denoise (OIDN) and NVIDIA OptiX.
Designed for production workflows with support for auxiliary buffers such as albedo, normal, and motion vectors (OptiX temporal).
- 🎯 High-quality denoising for CG renders
- ⚡ Dual backend support:
- Intel OIDN (CPU / GPU depending on build)
- NVIDIA OptiX (GPU accelerated)
- 🧠 Temporal denoising support (OptiX)
- 🎨 Auxiliary buffers:
- Albedo
- Normal
- Motion vectors (for temporal stability)
- 🔧 Configurable quality, device, and filtering modes
- Nuke 17.0v1
- Nuke 16.0v2
- Nuke 15.2v4
Other versions may be suported, however some manual fixing will be needed.
See BUILDING.md for information about installating.
All inputs at the moment require being shuffled into RGB space. You will also need to denoise and copy the Alpha over seperately.
| Input | Description |
|---|---|
color |
Main beauty render |
albedo |
Diffuse albedo pass (optional) |
normal |
World or camera space normals (optional) |
motion |
Motion vectors (required for temporal OptiX) |
| Setting | Type | Description |
|---|---|---|
Engine |
Enumeration | The technique used for denoising |
| Setting | Type | Description |
|---|---|---|
Device |
Enumeration | The hardware backend used for OIDN denoising |
Filter |
Enumeration | The filter method used for OIDN denoising |
Quality |
Enumeration | Image quality |
Mode |
Enumeration | Input image encoding. Use HDR unless the input is sRGB (2.2 gamma) or linear, in which case use sRGB. Output follows the same encoding |
Clean Aux |
Boolean | Denoise auxiliary features (albedo, normal) |
Input Scale |
Float | Scales input values before filtering without affecting output range. Useful for mapping HDR values to expected ranges |
Directional |
Boolean | Indicates input contains normalized directional lightmap coefficients (typically in [-1, 1]) |
| Setting | Type | Description |
|---|---|---|
Model |
Enumeration | The method used for OptiX denoising. Temporal mode requires motion vectors |
Blend |
Float | Denoising strength. 1.0 = fully denoised |
If you come across any Issues / Bugs, or are able to build the plugin for other versions of Nuke, please Create an Issue.
This project was created under the MIT license.
Special thanks to https://github.com/mateuszwojt/NukeCGDenoiser and https://github.com/twins-annning/NukeOIDNDenoise for the launching off point for this project. Without your code I definitely woudn't have made it this far.