View a markdown version of this page

SideFX Houdini - AWS Deadline Cloud
Support overviewHoudini version compatibilityDeadline Cloud Conda ChannelGetting startedInstallationUsing the Houdini submitterTroubleshootingAdvanced configurationsHoudini render enginesOpen source resources

SideFX Houdini

SideFX Houdini is a 3D procedural software for modeling, rigging, animation, VFX, look development, lighting, and rendering in film, TV, advertising, and video game pipelines. Houdini is fully supported by Deadline Cloud with comprehensive integration including submitters, conda packages, and an adaptor for increased rendering performance. This guide provides step-by-step instructions for using AWS Deadline Cloud with Houdini to render your projects faster by distributing rendering tasks across multiple machines.

Support overview

Houdini is supported by the following components:

  • Submitter: Integrated render output node (ROP) for direct job submission from Houdini with automatic scene and asset detection.

  • Conda packages: Deadline Cloud for automatic installation on service-managed fleets.

  • Adaptor: Middleware for efficient rendering with sticky sessions and additional monitoring.

  • Cross-platform compatibility: Submitter support for Windows, macOS, and Linux with worker support for Windows and Linux with automatic path mapping.

Houdini version compatibility

The following table shows current support levels for Houdini versions:

Major Version Submitter Support Conda Support Render Engines Usage-Based Licensing
19.0 Windows, macOS, Linux Linux Mantra, Karma CPU, Karma XPU Usage-based licensing available
19.5 Windows, macOS, Linux Linux Mantra, Karma CPU, Karma XPU Usage-based licensing available
20.0 Windows, macOS, Linux Linux Mantra, Karma CPU, Karma XPU Usage-based licensing available
20.5 Windows, macOS, Linux Linux Mantra, Karma CPU, Karma XPU Usage-based licensing available
21.0 Windows, macOS, Linux Linux Mantra, Karma CPU, Karma XPU Usage-based licensing available

Deadline Cloud Conda Channel

The following table lists all conda packages applicable to Houdini available to Service-managed fleets in the deadline-cloud conda channel:

OS Package Version Notes
Linux houdini 19.0 Includes Mantra and Karma renderers
Linux houdini 19.5 Includes Mantra and Karma renderers
Linux houdini 20.0 Includes Mantra and Karma renderers
Linux houdini 20.5 Includes Mantra and Karma renderers
Linux houdini 21.0 Includes Mantra and Karma renderers
Linux houdini-openjd Includes the Houdini Adaptor

Getting started

To use Houdini with Deadline Cloud:

  1. Create a service-managed fleet and associate it with a queue. Your queue must be set up with a queue environment that supports the deadline-cloud conda channel. For more information, see Creating a queue environment.

  2. Install the Deadline Cloud monitor and Houdini submitter on your artist workstation using the Deadline Cloud Submitter and monitor installers. For more information, see Set up your workstation.

  3. Submit your job directly from Houdini using the integrated submitter to the queue.

  4. Monitor the job and download the output using the Deadline Cloud monitor.

Installation

To install the Deadline Cloud for Houdini submitter, you need:

  • A Windows, macOS (arm64), or Linux workstation.

  • A supported version of Houdini.

Installing the submitter

To install the submitter

  1. Download the Deadline Cloud submitter installer.

  2. Run the installer.

    • When prompted, select each version of Houdini you want to use the submitter with.

  3. Launch Houdini.

The Deadline Cloud submitter is automatically available as a render output (ROP) node.

Note

The submitter installer is available for Windows, macOS, and Linux. For manual installation, see the manual installation instructions in the GitHub repository.

Verifying the submitter is installed correctly

  1. Open Houdini.

  2. In the Network Editor, choose the /out network.

  3. Open the context menu (right-click or press Tab) and search for deadline.

  4. Choose Deadline Cloud to create a new node.

Using the Houdini submitter

The Deadline Cloud for Houdini submitter is a node that accepts a render output (ROP) node as input. You can configure and submit your job through this node. When you submit a job, it includes steps for each ROP in the graph.

Submitting a job from Houdini

To use the Deadline Cloud for Houdini submitter, you need:

  • A profile to submit to Deadline Cloud with.

  • An Deadline Cloud farm and queue to submit to.

To submit a job from Houdini to Deadline Cloud

  1. In the Network Editor, choose the /out network.

  2. Open the context menu (right-click or press Tab) and search for deadline to create an Deadline Cloud node.

  3. Connect the output of a ROP to the input of the Deadline Cloud node.

    • When you connect a node to the Deadline Cloud node, the submitted job renders the input ROP and all ROPs in its graph.

  4. Select the Deadline Cloud node.

  5. Use the options in the node editor to configure your job. See Houdini-specific settings for information about what each option does.

  6. (Optional) To export a job's associated files to your job history directory without submitting it, choose Export Bundle.

  7. Choose Submit to send your job to Deadline Cloud.

Houdini-specific settings

The Job-specific settings tab of the Deadline Cloud node provides options specific to Houdini jobs.

  • Submit Dependencies as Separate Steps - Split the ROP graph into separate rendering steps for easier monitoring and debugging. When enabled, each connected render node becomes its own step in the job.

  • Include Adaptor Wheels - Enable custom builds of the adaptor (called wheels) that change rendering behavior. When enabled, you can specify a directory containing adaptor wheels. You can build adaptor wheels by running the build_wheels.sh script in the GitHub repository.

  • Adaptor Wheels - Specify the directory path containing custom adaptor wheels (only available when Include Adaptor Wheels is enabled).

  • Automatically unlock ROPs - Automatically unlock dependency ROPs during submission. Locked ROPs use existing outputs and won't re-render, which can block dependencies from re-rendering.

  • Automatically parse scene (.hip) references - Automatically discover and attach the job's input and output file names and directories based on the ROP graph during job submission.

  • Automatically save scene (.hip) file - Automatically save the scene (.hip) file to $HIP when submitting a job.

For information about the other submitter options, see the Deadline Cloud guide for using a submitter.

Overriding the render strategy for Deadline Cloud jobs

For many types of nodes, frames can be rendered independently and in any order. For others like simulations, each frame depends on the result of the previous frame and must be rendered sequentially. The submitter chooses a rendering strategy for each node based on its type, but also allows you to override the default.

Parallel vs. sequential rendering

For parallel rendering, each frame has its own task, and the tasks are distributed across available workers. For sequential rendering, all frames for a node are rendered in a single task running on a single worker.

By default, if a node is a geometry node with Initialize Simulation OPs enabled, it renders sequentially. Otherwise the node renders in parallel.

Adding a render strategy parameter

You can override the render strategy by creating a deadline_cloud_render_strategy parameter on your render node (for example, Mantra or Karma) with a value of either SEQUENTIAL or PARALLEL.

To override render strategy by adding a parameter

  1. Open the context menu for a node in the /out network (right-click).

  2. Choose Parameters and Channels, Edit Parameter Interface.

  3. Under Create Parameters, By Type, choose Ordered Menu.

  4. Add an ordered menu to Existing parameters by selecting the right arrow next to the Create Parameters column.

  5. Select the new parameter under Existing Parameters, then edit its configuration under Parameter Description:

    • In the Parameter tab:

      • For Name, enter deadline_cloud_render_strategy.

      • For Label, enter Deadline Cloud Render Strategy.

    • In the Menu tab, add menu items for:

      TokenLabel
      SEQUENTIALSequential
      PARALLELParallel

  6. Choose Accept.

Now in the parameter editor for your node, you can use the Deadline Cloud Render Strategy menu to specify submitter behavior.

Husk rendering and USD workflows

The following sections describe current limitations of USD export workflows in the Houdini submitter and an alternative example job bundle for rendering exported USD scenes with Husk.

USD export workflow support

The Deadline Cloud submitter for Houdini does not currently have built-in support for USD export workflows.

You cannot use the submitter node to create a single job that will export a USD scene from Houdini and then call Husk standalone to render without consuming a Houdini Engine license.

Alternative: example Husk job bundle

Deadline Cloud provides an example Husk job bundle that enables USD export rendering workflows outside of the Houdini submitter. You will need to export the USD scene yourself separately from Houdini before using the example job bundle.

The Husk example job bundle:

  • Allows direct submission of USD scenes for rendering using Husk and a chosen Hydra render delegate without launching Houdini and consuming a Houdini engine license during the render.

  • Automatically introspects USD files to find any file dependencies within to attach using job attachments.

  • Provides a simple GUI for configuration of common Husk settings and submission.

Prerequisites

Before using the Husk example job bundle, you need:

  • A scene exported to USD format.

  • The Deadline Cloud CLI installed and configured.

  • A git clone of the deadline-cloud-samples repository.

  • The Hydra render delegate available on the worker nodes.

    • Karma is included with Houdini. If you want to use other Hydra render delegates, you must provide them on the worker. See the deadline-cloud-samples repository for example conda packages for V-Ray and Redshift as one option to make them available on the worker nodes.

Using the Husk example job bundle

To use the Husk example job bundle

  1. Submit the bundle using the Deadline Cloud CLI:

    deadline bundle gui-submit ./deadline-cloud-samples/job_bundles/houdini_husk_usd_render

  2. Configure your USD file, output settings, frame range, and any other applicable settings to submit.

Husk example job bundle GUI interface.
Additional resources

Troubleshooting

The following sections describe common errors and questions you might encounter when using the Deadline Cloud submitter for Houdini, and how to resolve them.

Why do I get "incomplete asset definitions" errors while rendering?

Jobs from this submitter that run in your farm may produce errors in the logs that look like:

The following node types are using incomplete asset definitions:
  Driver/deadline_cloud

These errors are safe to ignore. The Deadline Cloud submitter exists as a node in your Houdini scene. When a worker in your farm loads the scene, the scene still contains the Deadline Cloud node, but the worker may not have the submitter installed. Because the worker does not have the files needed to run the Deadline Cloud node, it logs "incomplete asset definition" errors. The Deadline Cloud node itself is not rendered as part of the job, so these errors can be ignored.

Does the Deadline Cloud submitter support USD export render workflows using Husk?

The Houdini submitter does not directly support export workflows using Husk at this time. Jobs created through the submitter always run the adaptor which uses hython and therefore a Houdini engine license for the duration of the render. If you want to render an exported USD scene using just Husk and a Hydra render delegate, you can use an example job bundle. This approach is useful to render USD scenes with only a render license (for example, Karma) without needing a Houdini engine license for the entire render. For more information on rendering USD scenes with Husk on Deadline Cloud, see Husk rendering and USD workflows.

Advanced configurations

Using unsupported versions

Deadline Cloud only supports and tests the workstation and worker software versions in the table above. When using the submitter, the worker will attempt to install the same version as used on the workstation. This may fail if the workstation version of Houdini does not appear in the version table above.

If you require an unsupported version of Houdini, you have the following options:

  • When submitting the job from Houdini, you may override the CondaPackages queue parameter to specify a supported version to use on the worker (for example, houdini=21.0, houdini-openjd=*). This may or may not work, depending on the features used by your scene and how Houdini works with scenes from your workstation version.

  • You may build a custom conda recipe and channel for your desired version to be installed on the worker. Use the conda recipe for a supported version linked below as a starting point, and package your desired version in a custom conda channel. For more information about creating custom conda channels, see Creating custom conda channels.

Houdini render engines

Houdini supports multiple render engines that are compatible with Deadline Cloud:

Render Engine Description GPU Support
Karma CPU Modern USD-based renderer (CPU variant) CPU-based
Karma XPU Modern USD-based renderer (GPU variant) GPU accelerated
Mantra Traditional Houdini renderer CPU-based
Arnold Third-party Monte Carlo ray tracer GPU/CPU hybrid
V-Ray Third-party photorealistic renderer GPU/CPU hybrid
Redshift GPU-accelerated renderer GPU optimized

These render engines are automatically detected and configured by the Houdini integrated submitter and usage is automatically licensed. The submitter maintains dependency trees between connected render output nodes (ROPs).

Open source resources

The submitter and adaptor are open source and available on GitHub. Houdini Conda recipes are available on GitHub for supported versions.