What Is OpenImageDenoise in Blender?

OpenImageDenoise (OIDN) is an AI-accelerated denoising filter integrated into Blender’s Cycles render engine designed to remove visual noise from ray-traced images. This article explains what OpenImageDenoise is, how the underlying technology functions, why it is essential for optimizing render times, and how to configure it directly in your Blender render settings and compositor.

Understanding OpenImageDenoise

Developed by Intel, OpenImageDenoise is an open-source library of high-performance, deep-learning-based denoising filters. In ray-traced rendering like Cycles, noise appears as grainy artifacts caused by an insufficient number of light ray samples. OIDN uses artificial intelligence trained on massive datasets of noisy and clean images to recognize and remove this noise, reconstructing high-quality, final-frame images from low-sample renders.

Unlike hardware-exclusive denoisers like NVIDIA OptiX, OpenImageDenoise supports standard CPUs and a broad range of modern GPUs across different manufacturers, making it widely accessible.

Key Advantages in Cycles

  • Drastically Reduced Render Times: Because the denoiser effectively cleans up heavy noise, artists can lower the render sample count significantly without sacrificing output quality.
  • Superior Detail Preservation: OIDN excels at preserving fine textures, geometric edges, and subtle lighting transitions compared to traditional blurring filters.
  • Cross-Platform Compatibility: It works reliably across Intel, AMD, Apple Silicon, and NVIDIA hardware configurations.

How to Enable OpenImageDenoise in Render Settings

You can enable OIDN directly in the Blender render properties for real-time viewport previews or final production renders.

  1. Switch your render engine to Cycles in the Render Properties tab.
  2. Scroll down to the Sampling panel and expand the Denoise subpanel.
  3. Check the box to enable denoising for Render (for final outputs) or Viewport (for real-time previewing).
  4. Set the Denoiser dropdown menu to OpenImageDenoise.

Configuring OIDN Options

  • Passes (Input Data): Choose the auxiliary data passed to the AI.
    • None: Uses only the noisy RGB image.
    • Albedo: Uses color data to help distinguish surface colors from lighting noise.
    • Albedo + Normal (Recommended): Provides surface color and geometric surface normal data, yielding the sharpest and most accurate results.
  • Prefilter: Controls how auxiliary data is cleaned before denoising.
    • Accurate: Yields the highest quality with maximum detail retention.
    • Fast: Reduces processing time slightly, ideal for quick iterations or lower-spec systems.
    • None: Skips auxiliary prefiltering.

Using OpenImageDenoise in the Compositor

For maximum flexibility, OIDN can also be applied as a post-processing step in Blender’s Compositor rather than baking it directly into the render:

  1. Go to the View Layer Properties tab and check Denoising Data under the Passes panel.
  2. Open the Compositor workspace and check Use Nodes.
  3. Add a Denoise node (Shift + A > Filter > Denoise).
  4. Set the denoiser mode to OpenImageDenoise.
  5. Connect the Noisy Image, Denoising Albedo, and Denoising Normal outputs from the Render Layers node into the corresponding inputs of the Denoise node.
  6. Connect the output of the Denoise node into your Composite or Viewer node.

Using the Compositor method preserves the raw render output, allowing you to tweak denoising settings or switch denoising passes without re-rendering the 3D scene.