NVIDIA OptiX denoiser

On this page

Overview
How to

Overview

NVIDIA’s OptiX AI-Accelerated Denoiser uses deep learning to remove noise and “fireflies” from a rendered image. It is useful for getting a higher-quality preview of what a converged image will look like so you can make lighting decisions faster.

The NVIDIA OptiX denoiser requires an NVIDIA GPU with Kepler architecture or better.
The denoiser is not available on macOS.

How to

To... Do this

To...	Do this
Use the denoiser in the Render View	Turn on the Denoise button in the toolbar at the top of the render view. (This button is disabled if the plugin is not installed.) In the Render View, the denoiser always uses the “basecolor” image plane (if it exists) as the albedo input.
Use the denoiser in Mantra	In the parameter editor for the Mantra render node, click the Images tab, then click the Output sub-tab. Click the drop-down menu button next to the Pixel filter field and choose “NVIDIA OptiX Denoiser” (sets the field to `denoise optix`). In the Pixel filter field you can optionally specify an image plane to use as the albedo input to the denoiser with the `-a` option, for example: denoise optix -a mydiffuse
Choose which GPU to use	If you have more than one GPU in your computer, you can specify which one to use for denoising. Run the following in command-line to see OptiX-capable devices listed in order: > hgpuinfo -o Set the `$HOUDINI_NVIDIA_OPTIX_DEVICENUMBER` environment variable to the desired device ID, starting from `0` (for example, `0` to use the first device, `1` to use the second device).

Use the denoiser in the Render View

Turn on the Denoise button in the toolbar at the top of the render view. (This button is disabled if the plugin is not installed.)

In the Render View, the denoiser always uses the “basecolor” image plane (if it exists) as the albedo input.

Use the denoiser in Mantra

In the parameter editor for the Mantra render node, click the Images tab, then click the Output sub-tab. Click the drop-down menu button next to the Pixel filter field and choose “NVIDIA OptiX Denoiser” (sets the field to denoise optix).

In the Pixel filter field you can optionally specify an image plane to use as the albedo input to the denoiser with the -a option, for example:

denoise optix -a mydiffuse

Choose which GPU to use

If you have more than one GPU in your computer, you can specify which one to use for denoising.

Run the following in command-line to see OptiX-capable devices listed in order:
```
> hgpuinfo -o
```
Set the $HOUDINI_NVIDIA_OPTIX_DEVICENUMBER environment variable to the desired device ID, starting from 0 (for example, 0 to use the first device, 1 to use the second device).

Rendering

Mantra user guide

Basics

Working with render nodes
Render region tool

Lets you drag a box in the viewer to render an automatically updating preview of that part of the view.
Setting up cameras
Expressions in file names

How to use variables and expressions in file path fields to generate numbered and unique filenames.
Creating a flipbook animation

Renders the contents of the viewer as images, allowing you to play them back at full speed.

Lighting

Lights and shadows
Environment lighting
Light bank

The light bank is a data tree view that lets you set various parameters on all lights in a scene through a single interface.
Linear lighting and color
Instancing lights

You can instance lights onto points at render time just like instancing geometry.
Light path expression

Next steps

Rendering workflow tips

Tips on how to integrate Houdini and Mantra into a production pipeline efficiently.
Understanding Mantra raytraced shading

Explains the background details of how Mantra uses raytracing to shade objects in the scene.
Render quality and speed
Rendering volumes
Subdivision surfaces and curves
Rendering as part of a workflow
Deep camera maps
/render/cryptomatte
IFD Workflows

A quick overview of how to reduce the size of IFD files when rendering with Mantra. How to use procedurals, Alembic archives, and packed primitives to make IFDs smaller, and how IFDs work with HQueue.
Saving IFD geometry

When you generate an IFD file for rendering, you can choose whether to inline the geometry in the IFD file, or store the geometry in external files.
Render with a large number of polygons
How to render wireframes
HQueue

HQueue is Houdini’s free distributed render and simulation manager.
NVIDIA OptiX denoiser

Guru-level

Other renderers