The pdg.Dictionary constructor and pdg.Dictionary.fromDictionary method, now take an optional argument that specifies the default array datatype when constructing a pdg.Dictionary from a Python dictionary. The default datatype is used when an empty array value is found in the source dictionary, which would normally make it impossible to infer the type of the array.

Fixed the documentation for hou.removeContextOption.

A Switch LOP in "Warn and output empty stage" mode used to cause the next LOP (connected to the Switch LOP's output) to error. This no longer occurs, and the Switch LOP outputs an empty stage as expected.

The default value for curve/point widths was changed to be 1.0 in karma to match Hydra 1 defaults. This should only affect scenes which didn't specify the widths explicitly.

Fixed a crash in the Edit Properties From Node LOP when the source LOP node path didn't point to an existing node.

The Asset Catalog pane now has a button to switch to a "tree view" of the current catalog data source.

Reworded the prompts in the Dynamic Motion tool to be more succinct.

Reworded the prompts in the Dynamic Motion tool to be more succinct.

Added support for creating HAPI materials with COPernicus nodes, and extracting them to an image in memory or to a file.

mix() is properly syntax-colored as a built-in function in OpenCL contexts.

The mpmsolver SOP will no longer error out on colliders containing shear transformations. A warning will be displayed to the user instead and the shear component will be ignored.

The mpmsolver SOP will no longer error out on colliders containing shear transformations. A warning will be displayed to the user instead and the shear component will be ignored.

• Added a new plugin settings "Enable Editable Node Tools",that can be used to change the default value used for itscorresponding property when instantiating HDA.

• Added shared memory session option.

• Added a new plugin settings "Enable Editable Node Tools",that can be used to change the default value used for itscorresponding property when instantiating HDA.

• Added shared memory session option.

Fixed a bug where PDG/TOPs failed to retrieve logs from tasks submitted to Deadline.

Fixed a bug where PDG/TOPs failed to retrieve logs from tasks submitted to Deadline.

Fixed a bug where PDG/TOPs failed to retrieve logs from tasks submitted to Deadline.

Render-time procedurals based on the HoudiniProceduralAPI can now be ordered and chained together (e.g., the processed/generated output of one procedural can be used as the input to another)

Added parameters to the Python Processor that allow it to filter and ignore upstream failed dependencies. These options were already available via the pdg.NodeOptions configuration object in Python, but previously weren't promoted up to parameters on the Python Processor itself.

Added the ML Example Raw render node along with the ROP ML Example Raw Output geometry node. These nodes allow a data set consisting of packed examples to be written out more efficiently than before. Previously, such data sets would be written out using the ML Example Output node, which would temporarily store a serialized data set as geometry before actually writing it to disk. The new nodes write out the data set more directly and more efficiently, skipping the intermediate geometry representation. The ML Example Output node has been modified to use ROP ML Example Raw Output internally, allowing existing setups to benefit from this optimization without needing to modify them. ML Example Output remains fully compatible with its previous implementation.

Fixed issues where mouse events being consumed by Qt widgets would result in obsolete mouse positions in UI events.

Fixed a bug that prevented saving shelf edits to disk in Houdini Core.

Fixed an issue with the Crowd Trigger DOP's "RBD Impact Data" mode which could cause slow performance.