Fixed a bug where enabling both lens shader and indirect guiding would cause the lens shader to reuse sample index between primary passes, causing jaggy appearance due to stuck pixel filter jitter.

Fixed a bug where enabling both lens shader and indirect guiding would cause the lens shader to reuse sample index between primary passes, causing jaggy appearance due to stuck pixel filter jitter.

Fixed the RigTree panel that wasn't updating to the current node on creation of the panel.

Added Qt5QuickControls2, Qt5QuickShapes and Qt5QuickTemplates2 libraries to Houdini's Qt distribution. These libraries enable users to run basic QML and QtQuick code.

OpenCL device selection in Miscellaneous preferences now shows the index of each option. Also, hgpuinfo will report the UUID of each compute device that supports the required extension.

When doing renders to a LOP Viewer, include AOVs linked to render products thof any type, not just "raster" render products. This allows renderer- specific render product AOVs to be accessible through the viewport.

Re-apply a change that allows Houdini to generate unique selection strings for nested point instancer instances by setting the environment variable HOUDINI_PROTOPATHS_IN_INSTANCE_SELECTIONS=1.

Fix crash when using negative primitive indices with rig::PrimConstraint

Fix bug where lights could still affect viewport after being deleted

Improve performance in LOPs viewport

Fix bug where the APEX Add Groom SOP would not get created by default as an oval node shape.

Fixed random crashes in the Houdini (main) Python 3.11 build due to incompatibilities with Houdini's PySide2 5.15.2 package.

Bumped up Houdini's PySide2 to 5.15.15, which has complete Python 3.11 support.

Packages - Add HOUDINI_SCRIPT_PATH legacy support

• Update hou.ui.loadPackage/unloadPackage/loadPackageArchive to supportscripts located in 'scripts/python` sub-folders.

Have ML Regression Train use numpy==1.26.0 instead of numpy==2.1.0. This gets rid of warnings and prevents potential incompatibility issues with torch.

Ensure that ML Regression Train is capable of GPU training on windows machines by obtaining the torch package from download.pytorch.org.

Mantra will now properly save out cryptomatte metadata when saving multi-part OpenEXR images using the OIIO image driver.

Fixed a possible precision issue when computing instance motion segments. This might have caused over (or under) blurring when the time segments on the stage didn't align with shutter times.

The context menu for a USD prim with no variant sets will now look up the scene graph hierarchy for any ancestor prims with variant sets, and allow setting those variants from the menu. This makes it much easier to choose variants on an asset by bringin up the menu for any prim within the asset.

The context menu for a USD prim with no variant sets will now look up the scene graph hierarchy for any ancestor prims with variant sets, and allow setting those variants from the menu. This makes it much easier to choose variants on an asset by bringin up the menu for any prim within the asset.

Using the scene graph tree to make a prim visible when an ancestor prim is marked as invisible now works as expected. Ancestor prims are marked visible, and the invisibility marker is moved onto the child prims (except along the path of prims we are marking as visible).

Using the scene graph tree to make a prim visible when an ancestor prim is marked as invisible now works as expected. Ancestor prims are marked visible, and the invisibility marker is moved onto the child prims (except along the path of prims we are marking as visible).

Fixed a potential crash in the Crowd MotionPath Edit SOP.

Fix crash in Texture Optical Flow and Texture Feature SOPs when input is a volume with multiple primitives.

Fix regression in the Rig Pose SOP's viewport tool where handles were rendered in the wrong space whenever its owner Object node was transformed.

Fix regression in the Rig Pose SOP's viewport tool where handles were rendered in the wrong space whenever its owner Object node was transformed.