When set to Path Traced, maximum of 1 indirect ray is generated per bounce. When set to Automatic, the number of indirect rays is calculated based on initial noise estimate, target noise threshold, and the maximum number of camera rays. Also note that under Automatic mode, number of samples for direct lighting is adjusted based on noise estimate as well.

Whether Karma should perform uniform sampling of lights or whether rendering should use the light tree. The light tree can be significantly faster for scenes that have large numbers of lights.

Some lights cannot be added to the light tree, and will all be sampled by Karma:

• Dome Lights

• Distant Lights

• Point Lights

• Lights with Light Filters

• Lights with shaping controls (i.e. spot lights)

This is a global control to improve sampling quality for all lights. This acts as a multiplier on the individual light quality controls. Increasing the quality will improve direct light sampling as well as shadows/occlusion.

The number of transparent samples to be shaded as a ray travels through partially opaque objects. Increasing this value will result in less noise in partially opaque objects and is generally less costly than increasing Pixel samples, Volume Step Rate, or Min and Max ray samples. This parameter will not have any effect on noise from Indirect Sources however.

How finely or coarsely a volume is sampled as a ray travels through it. Volumetric objects are made up of 3d structures called Voxels, the value of this parameter represents the number of voxels a ray will travel through before performing another sample.

The default value is 0.25, which means that every one of every four voxels will be sampled. A value of 1 would mean that all voxels are sampled and a value of 2 would mean that all voxels are sampled twice. This means that the volume step rate value behaves in a similar way to pixel samples, acting as a multiplier on the total number of samples for volumetric objects.

Keep in mind that increasing the volume step rate can dramatically increase render times, so it should only be adjusted when necessary. Also, while increasing the default from 0.25 can reduce volumetric noise, increasing the value beyond 1 will rarely see similar results.

The number of times diffuse rays can propagate through your scene.

Unlike the Reflect and Refract Limits, this parameter will increase the overall amount of light in your scene and contribute to the majority of global illumination. With this parameter set above zero diffuse surfaces will accumulate light from other objects in addition to direct light sources.

In this example, increasing the Diffuse Limit has a dramatic effect on the appearance of the final image. To replicate realistic lighting conditions, it is often necessary to increase the Diffuse Limit. However, since the amount of light contribution usually decreases with each diffuse bounce, increasing the Diffuse Limit beyond 4 does little to improve the visual fidelity of a scene. Additionally, increasing the Diffuse Limit can dramatically increase noise levels and render times.

The number of times a ray can be reflected in your scene.

This example shows a classic “Hall of Mirrors” scenario with the subject placed between two mirrors.

This effectively creates an infinite series of reflections.

From this camera angle the reflection limits are very obvious and have a large impact on the accuracy of the final image. However, in most cases the reflection limit will be more subtle, allowing you to reduce the number of reflections in your scene and optimize the time it takes to render them.

Remember that the first time a light source is reflected in an object, it is considered a direct reflection. Therefore, even with Reflect Limit set to 0, you will still see specular reflections of light sources.

This parameter control the number of times a ray be refracted in your scene.

This example shows a simple scene with ten grids all in a row.

By applying a refractive shader, we will be able see through the grids to an image of a sunset in the background.

From this camera angle, in order for the image to be accurate, the refraction limit must match the number of grids that that are in the scene. However, most scenes will not have this number of refractive objects all in a row and so it is possible to reduce the refract limit without affecting the final image while also reducing the time it takes to render them.

Keep in mind that this Refract Limit refers to the number of surfaces that the ray must travel through, not the number of objects.

Remember that the first time a light source is refracted through a surface, it is considered a direct refraction. Therefore, even with Refract Limit set to 0, you will see refractions of Light Sources. However, since most objects in your scene will have at least two surfaces between it and the light source, direct refractions are often not evident in your final render.

The number of times a volumetric ray can propagate through a scene. It functions in a similar fashion to the Diffuse Limit parameter.

Increasing the Volume Limit parameter will result in much more realistic volumetric effects. This is especially noticeable in situations where only part of a volume is receiving direct lighting. Also, in order for a volumetric object to receive indirect light from other objects, the Volume Limit parameter must be set above 0.

With the Volume Limit set to values above zero, the fog volume takes on the characteristic light scattering you would expect from light traveling through a volume. However, as with the Diffuse Limit, the light contribution generally decreases with each bounced ray and therefore using values above 4 does not necessarily result in a noticeably more realistic image.

Also, increasing the value of this parameter can dramatically increase the amount of time spent rendering volumetric images.

The number of times a SSS ray can propagate through a scene. It functions in a similar fashion to the Diffuse Limit parameter.

The maximum value a shading sample is allowed to contribute to an LPE image plane to reduce appearance of “fireflies” caused by undersampling of extremely bright light sources. Note that reducing this value can result in an overall reduction in the amount of light in your scene.

When enabled, indirect bounces use Color Limit value and Indirect Color Limit parameter is ignored.

Color limit applied to indirect bounce only. Note that this parameter is ignored unless Shared Color Limit toggle is disabled.

Depth at which indirect rays start to get stochastically pruned based on ray throughput.

Enable depth of field rendering.

Whether motion blur should be on or off for objects that don’t explicitly have an opinion. If this is On by default, the parameters below let you set the defaults.

The number of samples to compute when rendering transformation motion blur over the shutter open time. The default is 2 samples (at the start and end of the shutter time), giving one blurred segment.

If you have object moving and changing direction extremely quickly, you might want to increase the number of samples to capture the sub-frame direction changes.

In the above example, it requires 40 transformation samples to correctly render the complex motion that occurs within one frame. (This amount of change within a single frame is very unusual and only used as a demonstration.)

Transformation blur simulates blur by interpolating each object’s transformation between frames, so it’s cheap to compute but does not capture surface deformation. To enable blurring deforming geometry, increase karma:object:geosamples.

The number of sub-frame samples to compute when rendering deformation motion blur over the shutter open time. The default is 1 (sample only at the start of the shutter time), giving no deformation blur by default. If you want rapidly deforming geometry to blur properly, you must increase this value to 2 or more. Note that this value is limited by the number of sub-samples available in the USD file being rendered. An exception to this is the USD Skel deformer which allows.

“Deformation” may refer to simple transformations at the Geometry (SOP) level, or actual surface deformation, such as a character or object which changes shape rapidly over the course of a frame.

Objects whose deformations are quite complex within a single frame will require a higher number of Geo Time Samples.

Deformation blur also lets you blur attribute change over the shutter time. For example, if point colors are changing rapidly as the object moves, you can blur the Cd attribute.

Increasing the number of Geo Time Samples can have an impact on the amount of memory Karma uses. For each additional Sample, Karma must retain a copy of the geometry in memory while it samples across the shutter time. When optimizing your renders, it is a good idea to find the minimum number of Geo Time Samples necessary to create a smooth motion trail.

Deformation blur is ignored for objects that have Velocity motion blur turned on.

This parameter lets you choose what type of geometry velocity blur to do on an object, if any. Separate from transform blur and deformation blur, you can render motion blur based on point movement, using attributes stored on the points that record change over time. You should use this type of blur if the number points in the geometry changes over time (for example, a particle simulation where points are born and die).

If your geometry changes topology frame-to-frame, Karma will not be able to interpolate the geometry to correctly calculate Motion Blur. In these cases, motion blur can use a velocities and/or accelerations attribute which is consistent even while the underlying geometry is changing. The surface of a fluid simulation is a good example of this. In this case, and other types of simulation data, the solvers will automatically create the velocity attribute.

When this is set to “Velocity Blur” or “Acceleration Blur”, deformation blur is not applied to the object. When this is set to “Acceleration Blur”, use the karma:object:geosamples property to set the number of acceleration samples.

When defining motion blur on instances, the transform of each instance can be blurred in addition to any motion blur occurring on the prototype. This option controls how the instance will compute the motion blur of the transform on each instance. For example, when instancing prototypes to a particle system, you'd likely want to use velocity blur to compute motion blur (the transform on the prototype would be blurred by the velocity on the particles).

Velocity multiplier used to reduce or exaggerate amount of motion blur on volumes.

When rendering point clouds, they can be rendered as camera oriented discs, spheres or discs oriented to the normal attribute.

When rendering curves, they can be rendered as ribbons oriented to face the camera, rounded tubes or ribbons oriented to the normal attribute attached to the points.

USD supports Curve Basis types that may not be supported directly in Houdini. In some cases, you may want to override the Houdini curve basis. For example, if you have linear curves in Houdini, you may want to render them with a Bezier, B-Spline or Catmull-Rom basis. This menu will force Karma to override the basis that’s tied to the USD primitives.

Note that the topology of the curves must match the target basis. For example, when selecting any cubic curve basis, every curves must have at least 4 vertices. For the Bezier basis, curves must have 4 + 3*N vertices.

If enabled, geometry that are facing away from the camera are not rendered.

Increasing this value can make caustics less noisy at the cost of accuracy.

Finds RenderVar prims in this node’s second input and adds them to this stage, so they add to the list of render vars to generate. This allows other LOP nodes (such as Background Plate) to “offer” render vars related to that node to be generated.

Finds RenderProduct prims in this node’s second input and adds them to this stage, so they add to the list of products to generate. This allows other LOP nodes to “offer” products related to that node to be generated.

Specifies the distribution of samples over pixels. A box filter will distribute samples randomly over the interior of each individual pixel. A Gaussian filter will distribute samples in a disk around the pixel center, but with a Gaussian distribution (instead of a uniform distribution).

This is the size of the Pixel Filter. A Guassian filter with a filter size of 1.8 will be slightly less blurry than a Gaussian filter with a filter size of 2.0.

When this is on, the renderer creates additional AOVs specific to each tagged light. To manage LPE Tags for lights, use the LPE Tag LOP.

Choose a denoiser to run on the finished image output, or No Denoiser. This utility currently supports Intel Open Image Denoise (included with Houdini) and the NVIDIA OptiX Denoiser (must be installed separately). You must be on a supported platform and have the chosen denoising library installed for this to work.

The NVIDIA OptiX Denoiser only works with NVIDIA cards. It is now included with the NVIDIA driver (version 435 or later).

Some denoising libraries can use albedo to get a better sense of the image, guiding how and where it reduces noise.

Some denoising libraries can use normals to get a better sense of the image, guiding how and where it reduces noise.

???

Space-separated list of AOVs to run the denoiser on.

OCIO image filters can be added to various render vars/image planes.

Enables the OCIO image filter defined below.

The render var names to which the OCIO image filter will be applied.

Specify the OCIO color output space the image filter will apply.

In most cases this should be left at data. There may be some strange case where the source of the AOV is already in a defined color space, in which case you can specify the source space here.

Space-separated list of looks to apply to the finished image. A “look” is a named OCIO color transform, usually intended to achieve an artistic effect. See the OCIO documentation for more information.

A color limit clamps the value a shading sample is allowed to contribute to an LPE image plane, to reduce appearance of “fireflies” caused by undersampling of extremely bright light sources. This multiparm lets you apply separate limits to different sets of render vars.

Enables the color limit sample filter defined below.

A space-separated list of AOV names to which this color limit will be applied.

The maximum value a shading sample is allowed to contribute to these AOVs.

What to do if the aspect ratio of the output image (Resolution width divided by height) doesn’t match the aspect ratio of the camera aperture (controlled by attributes on the camera). This allows a standard renderer to do something reasonable when you switch between cameras.

Directs the renderer to only render within this window of the entire output image. You specify the window as minX, minY, maxX, maxY, where each number is a normalized value from 0 to 1. 0, 0 is the bottom left, 1, 1 is the top right, 0.5, 0.5 is the center, and so on. The default is 0, 0, 1, 1 (no cropping). Note that you can use negative values. For example, -0.1, -0.1, 1.1, 1.1 will give you 10% overscan on each side.

You can use this window to temporarily crop the render to a smaller region, for testing purposes.

Pixels are only rendered if they are fully inside the window.

The normalized coordinates map to the image after any adjustments by the Aspect ratio conform policy.

The aspect ratio (width/height) of image pixels (not the image itself). The default is 1.0, indicating square pixels.

The name of the person, department, or studio that created the image file. The node will set this field on the output image if the image format supports metadata (for example, .exr).

An arbitrary comment, for example a description of the purpose of the output image. The node will set this field on the output image if the image format supports metadata (for example, .exr).

The name of the computer that generated this the output file. The node will set this field on the output image if the image format supports metadata (for example, .exr).

The type of compression to apply to .exr output files.

If this is on, Karma does a pre-render to roughly estimate of the light in the scene, and uses that to guide indirect diffuse rays, rather than just relying on the BSDF sampling distribution. This can improve “difficult” lighting (for example, caustics, and mostly indirect lighting), but can make “easy” lighting noisier. Before using this, you can try rendering direct and indirect AOVs to see where the noise is. If the noise is mostly caused by the direct lighting, there’s no point in turning on path guiding.

Number of pixel samples to use for prepass render when Indirect Guiding is enabled. Recommended minimum is 10% of pixel samples under Path Traced convergence mode, or the same number of pixel samples under Automatic convergence mode. Grid-like artifacts can show up in the final render when this value is too low and not enough data is collected. It’s recommended to keep this value above 64.

When rendering, a Pixel Oracle tells karma which pixels need additional sampling and which pixels are converged. This parameter tells karma which oracle to use.

The minimum number of camera rays (primary samples) for each pixel.

The AOV to use to measure variance.

The amount of variance that triggers more rays.

Variance sampling involves some randomness. You can change this number to get slightly different results.

Whether to apply an OCIO transform to the pixels before measuring variance.

When OCIO Transform is Display View, the display to transform to before measuring variance.

When OCIO Transform is Display View, the view to transform to before measuring variance.

When OCIO Transform is Explicit, the color space to transform to before measuring variance.

When this is on, Karma will periodocially write out image tile data to a checkpoint file. If the process is terminated before completing the render, you can resume it by turning on Resume from Checkpoint and restarting.

When Output checkpoint files is on, the name of the checkpoint file to write to. The default ($HIP/render/$HIPNAME.$OS.$F4.checkpoint) puts the checkpoint file inside a render directory next to the current scene file, and includes the base name of the current scene file ($HIPNAME), this node’s name ($OS), and the render frame ($F) in the filename to help avoid two processes trying to use the same checkpoint file at the same time.

When Output checkpoint files is on, Karma waits this number of seconds between writing out a checkpoint file. The default is 60.

If this is on and you start a render and the renderer notices there is a valid checkpoint file, it will try to resume rendering from that checkpoint. If you want to restart the render from the beginning, you can turn this off or just delete the checkpoint file(s).

Karma breaks down an image into multiple buckets for rendering. This is the side length (in pixels) of the square bucket. The default is 32, specifying a 32 pixel x 32 pixel bucket. Threads operate at the bucket level, so it might be useful to lower the bucket size if there are only a few buckets that are particularly expensive. That way the expensive areas can be divided across more threads.

For example, if the image is mostly empty, but there’s a distant object that fits within single 32 x 32 bucket, then that object will only be rendered using 1 thread. If you switch to a 16 x 16 bucket, then the object might be split across 4 buckets and have 4 threads working on it.

Ideally changing the bucket size doesn’t change the results, but Karma measures variance across pixels within the current bucket, so if you set it to a low value, for example 4, Karma only has 4 x 4 = 16 pixels to look at, so Karma will tend to make very poor variance estimates. This can show up as black pixels, where pixel rendering terminated prematurely due to a bad variance estimate.

Specifies which buckets are rendered first. Values can be:

Note: When rendering to mplay, the user can click to focus on an area to render.

Whether to use a fixed size cache (karma:global:cachesize) or whether to use a proportion of physical memory (karma:global:cacheratio)

The proportion of physical memory Karma will use for its unified cache.

For example, with the default vm_cacheratio of 0.25 and 16 Gb of physical memory, Karma will use 4 Gb for its unified cache.

The unified cache stores dynamic, unloadable data used by the render including the following:

• 2D .rat texture tiles

• 3D .i3d texture tiles

• 3D .pc point cloud pages (when not preloaded into memory)

Note: This value is only used for off-line rendering, not IPR.

An explicit memory limit for the unified shading cache. This is deprecated in favor of using the Cache Memory Ratio.

Note: This value is only used for off-line rendering, not IPR.

Number of threads to use for rendering. 0 uses the number of processors. -1 uses the number of processors minus one.

Limit rendering to this number of seconds.

The directory where any required USD files will be saved. If this directory is in the $HOUDINI_TEMP_DIR, it will be deleted automatically when the render is complete.

Comma separated list of purposes to render (from geometry, guide, proxy, and render). The default is geometry,render.

This setting is defined by USD and Hydra to control for the number of iterations to use to render “refine-able” (subdivision) surfaces. Karma mostly ignores this setting, since it measures the surface in screen space to decide how many iterations it needs to render the surface smoothly. Instead Karma uses this setting more like a subdivision on/off switch. If this is set to Low or Medium, Karma renders the (chunky) hull. If it is High or Very High, Karma renders the (smooth) limit surface.

Other renderers (such as RenderMan or Storm) will treat this setting differently, using it to determine subdivision quality. If you need to render the same scene in Karma and another renderer, remember to use High or Very High to make sure Karma renders subdivision surfaces smoothly.

This file path is passed to the render command line as the resolver-context option. This argument is used to create an asset resolver context that helps the asset resolver find files while composing the USD stage. The default of this parameter is an expression that returns the corresponding parameter on the LOP Network that contains the selected LOP path

Use this multiparm to set global context options in effect during the render.

If you turn this off, when you start a render the node will generate the USD output files to render (in the directory specified in Advanced ▸ Options ▸ USD Output Directory), but will not actually run the renderer on them.

Enable this option to prevent layers authored above Layer Break nodes from being written to disk by this ROP. This allows a Layer Break node to dictate which portions of the LOP Network are to be saved. Disabling this option allows this behavior to be overridden, forcing the full stage authored by the LOP network to be written to disk. This may be used for debugging purposes, or to write a complete scene to disk for rendering.

Initialize DOP simulations before rendering.

Create intermediate parent directories for output files as needed, such as for generated images.

Print the number of bytes sent or received by the distributed simulation nodes during cooks triggered by this node.

This does not track network usage from, for example, saving a file to an NFS mount. It only tracks the network communication of distributed Houdini nodes.

Enabling this option will cause karma to stop the render with an error if it encounters a missing texture map.

Causes Houdini to freeze after starting the render process, and only resume when the renderer exits. (This is included mostly for parameter parity with the old Mantra render nodes. It may be useful to implement a crude form of dependency ordering when batch rendering.)

Increasing this value will cause more information to be printed out during rendering. To see render time and memory information, set the verbosity level to 1. To see more detailed information about the render, set this parameter to 3. Larger values print out progressively more information.

VEX profiling lets you analyze shader performance. Turning this on will slow down shading, especially when NAN detection is turned on.

As rendering progresses, print out progress using the Alfred progress format. For multi-frame renders, this progress will be cumulative across frames (i.e. if there are 4 frames, percent complete will be 25% after the first frame finishes).

Show the time next to the log entries output by the renderer.

How to show console output on Windows when the render is launched from a graphical application.

Normally husk will overwrite the log file on each run. This option will cause output to be appended to the text in existing files.

This is the file where normal messages should be saved.

This is the file to which error messages should be saved. If this is the same file specified for the Output Log, the output and error messages will be interleaved (in the same way they are written to the console).

Initialize DOP simulations before rendering.

When rendering to disk, open an MPlay window to interactively monitor progress.

This is a zoom factor applied to the monitor. The monitor will display images at a fraction of the full resolution while the image written to disk will be full resolution.