This plugin provides a free and open-source machine learning toolset for Houdini, targeting 3D artists and technical directors. It enables the integration of ML models, such as Stable Diffusion, directly within the Houdini environment for generative workflows.

How It Works

The plugin leverages Python and PyTorch to integrate ML models into Houdini's node-based workflow. It allows users to download and manage models from Hugging Face, directly applying them within custom Houdini Digital Assets (HDAs). This approach offers a powerful way to incorporate AI-driven content generation and manipulation into existing 3D pipelines.

Quick Start & Requirements

• Installation: Clone the repository, copy MLOPs.json to $HOUDINI_USER_PREF_DIR/packages/, update the path in MLOPs.json, launch Houdini, and use the "Install Dependencies" shelf button.
• Prerequisites: Git, Houdini.
• Dependencies: PyTorch (with CUDA support recommended for GPU acceleration), other Python libraries listed in requirements.txt.
• Setup: Requires downloading models (e.g., Stable Diffusion) via a shelf button.
• Documentation: https://github.com/Bismuth-Consultancy-BV/MLOPs

Highlighted Details

• Integrates Stable Diffusion and other Hugging Face models into Houdini.
• Supports GPU acceleration via PyTorch (CUDA/MPS).
• Provides shelf tools for dependency installation and model downloading.
• HDAs are distributed in expanded format; can be collapsed with hotl.

Maintenance & Community

• Developed by Ambrosiussen Holding and Entagma.
• Distributed by Bismuth Consultancy B.V.
• Community support via Discord.

Licensing & Compatibility

• Licensed and distributed by Bismuth Consultancy B.V.
• Users agree to the LICENSE and Terms of Service upon use. Specific license details are not fully elaborated in the README.

Limitations & Caveats

The plugin can conflict with other Houdini plugins like Renderman and Octane during dependency installation. Users may need to disable these temporarily. Corrupted model caches can cause errors, requiring manual deletion.