Page History

The render settings included in the Performance tab control how the image is rendered without affecting the appearance of the final result.

UI Path: ||Render Setup window|| > Performance tab (Renderer set to Corona)

Primary solver – Global illumination solver used for the primary (first) light bounce.

None – Disables GI, leaving only direct lighting.
Path Tracing – An accurate brute force solution.

Secondary solver – Global illumination solver used for the secondary (second, third, ...) light bounces.

None – Disables secondary GI, leaving only one bounce of GI.
Path tracing – An accurate brute force solution that is less efficient in scenarios with lots of indirect lighting (e.g. interior scenes), or generally scenes with complicated paths between lights and the rendered surfaces. It is suggested to use this solver mostly in exterior situations.
UHD Cache – A fast, interpolated, slightly less accurate solution. It works best in scenes with a lot of light bouncing, such as interior scenes, but may be inefficient in scenes with complex geometry and simple lighting, such as exterior foliage shots.
4K Cache – An experimental solver.

Lock sampling pattern – When enabled, the noise will be "locked" in place, both in still frames and animations, making it less visible. Disabling this option will produce a completely different noise pattern each frame, which is required by some third party de-noising applications (but not by the Corona denoiser). When rendering using Distributed Rendering or a render farm service, the noise might be different from frame to frame regardless of this option.

Conserve memory (slower) – When enabled, Corona consumes less memory at a cost of rendering performance. This option can decrease the peak memory usage by up to 30%. Consider turning this on if you are running out of memory during rendering.

Automated out-of-core textures – When enabled, Corona automatically offloads unused parts of textures to HDD to consume less memory. You can set details in the System Settings dialog.

Sampling Balance

GI vs. AA balance – Number of GI samples per each anti-aliasing sample. Higher values will make global illumination and light noise clean up faster, while lower values will make the antialiasing, motion blur, and depth of field noise clean up faster. Setting this value under 2 or above 64 is not recommended.

Light Samples Multiplier – Number of direct lighting samples per each GI sample. Increase this value in scenes with visible noise in direct lighting. Default value of 2.0 works well in most cases.

Speed vs. Accuracy Balance

Max Sample Intensity – Maximum brightness of the secondary GI samples. This is the tradeoff between rendering performance and physical accuracy. Lower values suppress noise, but produce a darker image with lower intensity of reflections and caustics. Higher values produce brighter reflections and caustics, but also more noise in the image. Default value of 20 works well in most scenes. Value 0 is a special case that enables an unbiased mode (this mode is not production-ready as it will never clear noise in some scenes).

Max ray depth – The maximum number of light bounces. Lowering this value slightly improves performance, increasing it slightly improves accuracy. However, the change in Corona is very small compared to other renderers, so keeping the default value is advised.

Displacement

Screen size (px) – When selected, displacement tessellation will be performed adaptively in the screen space. This is usually a lot more efficient. Size sets the number of pixels each tessellated triangle spans in the image. Lower values will improve displacement quality at the expense of memory usage and preprocessing time.

World size (units) – When selected, displacement tessellation will be performed absolutely in the world space.\r\n\r\nThis method is usually inefficient except for special cases (to prevent popping artifacts in animations). Size sets the maximum length of each tessellated triangle in world units. Lower values will improve displacement quality, at the expense of memory usage and render preprocessing time. Beware of setting this value too low and using this method on large objects, as this may consume all the available memory and result in a crash.

Interactive Rendering

Max passes – Limits the maximum number of passes rendered in the interactive rendering mode. When using this feature, your CPU and cooling will thank you for letting them rest. Note that for interactive render regions, this value is best left set to 0 to ensure that new or changed regions will render (otherwise they will not render if the max passes has already been reached).

Force Path Tracing – When on, Path Tracing will be always used as primary and secondary solver in interactive rendering, even when UHD Cache is selected for regular rendering. This keeps the interactive rendering responsive, avoiding the cost of UHD Cache precomputation.

Denoising during rendering – When set to another option than "None", the image is iteratively denoised during rendering. This replaces a noisy render preview with a noiseless, but lower-quality image which gets progressively refined as the rendering converges. It can help to get a quick estimate of the overall lighting in the scene. There are two denoising options available:

NVIDIA GPU AI – A very fast denoiser which is available only on NVIDIA GPUs when the optional "NVIDIA AI Denoiser" Corona component is installed.

INTEL GPU AI – Also known as Intel Open Image Denoise. A slower, but higher-quality denoiser.

Upscaling – When enable, NVIDIA AI upscaling is used for interactive rendering. Enable Upscaling automatically enables NVIDIA AI Denoiser.

Locked Resolution – From Corona 12 onward, you can lock the rendering resolution for interactive rendering in the VFB.

Enable – When enabled, the specified resolution will be used for interactive rendering in the VFB and prevent fitting the resolution to the whole VFB window during resizing.

Width – Specify the width of the locked resolution.

Height – Specify the height of the locked resolution.

Caustics Solver

Enable – Generates physically correct caustics at the cost of lower rendering performance. Note that to render refractive caustics you need to enable them in CoronaMtl.

0 objects excluded... button – Exclude/Include list of scene objects that can receive caustics.

"+" button – Adds the currently selected objects to the Caustic solver include/exclude list.

Only in caustics element – When checked, caustics will be rendered only in special CShading_Caustics render element. When unchecked, caustics will appear also in beauty, light select and all other applicable render elements.

Caustics from environment – If enabled, caustics solver will generate caustics from environment map lighting. This option does not influence caustics generation by sun.

Caustics in volumes – If set to true, the Caustic solver will be used to compute caustics in volumes.

Surface multiplier – Multiplies intensity of caustics on surfaces. Changing the default value of 1 leads to inaccurate (darker/brighter) surface caustics rendering.

Volume multiplier – Multiplies the intensity of caustics in volumes. Changing the default value of 1 leads to inaccurate (darker/brighter) surface caustics rendering.

DOF Highlights Solver [EXPERIMENTAL]

Enable – Enables the DOF Highlights Solver that improves speed of rendering of out of focus objects (at the cost of casting more rays).

Budget (%) – Increasing this number improves rendering of out of focus objects, but it leads to slower rendering of all other effects.

Reprojection count – For debugging purpose only. When highlight blurred by depth of field is found, it is projected back to the image. This number changes how many times the highlight is projected back.

Preset

Still frame (fast precomputation) – Configuration with short precomputation that works great in still frames, but might flicker in animations.

Animation (flicker-free) – Configuration with longer precomputation that aims at eliminating flickering in animations.

Precision – Quality-time tradeoff for the UHD Cache. Lower values result in fast precomputation but can increase bias and flickering. Higher values give more precise, flicker-free results with longer precomputation. High values should be avoided, as they will also lower the success rate of the cache and make it less efficient during rendering phase.

Save/Load

Precomputation – Determines how the UHD cache for rendering is obtained:

Calculate from scratch – The cache will be calculated independently for each frame in a precomputation pass.
Load from file – The cache will be loaded from the specified file instead of computing it. Good for animations as it prevents flickering.
Load from file + append – The cache will be first loaded from the specified file, and then the precomputation pass will run on top of it. The loaded and computed caches will be merged. Good for animation precomputation (when ran e.g. every 30 frames).

After render – When set to "Save to file", the UHD cache will be saved to the specified file after rendering, possibly overwriting it.

Filename – Specifies the path where the cache file is saved to or loaded from.

Preset

Still frame (fast precomputation) – Configuration with short precomputation that works great in still frames, but might flicker in animations.

Animation (flicker-free) – Configuration with longer precomputation that aims at eliminating flickering in animations.

Granularity – Increasing this value will lower memory usage, but it will result in more splotchy render.

Quality – Increasing this parameter will increase the precomputed GI quality, while increasing the precomputation time at the same rate.

Save/Load

Precomputation – Determines how the 4K cache for rendering is obtained:

Calculate from scratch – The cache will be calculated independently for each frame in a precomputation pass.
Load from file – The cache will be loaded from the specified file instead of computing it. Good for animations as it prevents flickering.
Load from file + append – The cache will be first loaded from the specified file, and then the precomputation pass will run on top of it. The loaded and computed caches will be merged. Good for animation precomputation (when ran e.g. every 30 frames).

After render – When set to "Save to file", the 4K cache will be saved to the specified file after rendering, possibly overwriting it.

Filename – Specifies the path where the cache file is saved to or loaded from.