N
About
Foundry Cattery is a free machine learning model library built specifically for visual effects and compositing artists working in Foundry's Nuke. Introduced alongside the Nuke 13 ML toolset — which includes CopyCat and the Inference node — Cattery solves a key friction point: while researchers frequently share powerful ML models on GitHub, converting and deploying them inside Nuke traditionally required engineering expertise. Cattery removes that barrier by providing a curated collection of pre-converted .cat model files that load directly into Nuke's Inference node with no additional technical work. Models hosted on Cattery cover a range of functionality including sequence-specific effects, image enhancement, and other visual processing tasks. Both Foundry-converted third-party PyTorch models and models trained natively with CopyCat are eligible for inclusion. The long-term vision is to open the platform to user submissions, enabling artists to share their own trained models with the global VFX community. Cattery is accessible to everyone regardless of studio size — whether you're an independent artist or part of a major production house. The initial release is download-only, hosted on the Foundry website, with no subscription or license fee required beyond having a compatible version of Nuke. It represents Foundry's commitment to democratizing machine learning in professional content creation and fostering collaboration between academic research and production pipelines.
Key Features
- Free .cat Model Library: Access a growing repository of pre-converted machine learning models in .cat format, ready to load directly into Nuke's Inference node at no cost.
- Inference Node Integration: All Cattery models are compatible with Nuke 13's Inference node, enabling artists to apply ML effects seamlessly within their existing compositing workflow.
- Third-Party PyTorch Model Support: Foundry converts popular third-party PyTorch models from the research community into .cat files, making cutting-edge AI accessible inside Nuke without manual conversion.
- CopyCat Compatibility: Models trained natively using CopyCat in Nuke can be exported and shared on Cattery, allowing artists to contribute their own custom-trained effects.
- Community Sharing (Roadmap): Foundry's long-term goal is to open Cattery to user submissions, creating a collaborative ecosystem where artists worldwide can contribute and discover new ML models.
Use Cases
- VFX artists applying pre-trained ML models to compositing shots in Nuke without needing a dedicated engineering team.
- Independent artists and small studios accessing the same AI-powered tools previously only available to large productions with technical staff.
- Studios sharing internally trained CopyCat models with the global VFX community via a common distribution platform.
- Researchers and ML engineers distributing their PyTorch models to production artists by converting them into the Nuke-compatible .cat format.
- Artists creating sequence-specific visual effects by downloading and applying ML models tailored to particular types of footage or styles.
Pros
- Completely Free: All models in the Cattery library are available at no cost, making professional-grade ML tools equally accessible to solo artists and large studios.
- No Engineering Expertise Required: Pre-converted .cat files eliminate the technical complexity of adapting research models, allowing artists to focus on creativity rather than infrastructure.
- Bridges Academia and Production: Cattery translates state-of-the-art ML research into production-ready VFX tools, closing the gap between what researchers share and what artists can actually use.
Cons
- Requires Nuke 13 or Later: Cattery models only work with Foundry's Nuke 13+ via the Inference node, limiting access to users on older versions or different compositing applications.
- Nuke Ecosystem Lock-In: The library is exclusively designed for Nuke, making it unavailable to artists using other compositing or VFX tools such as After Effects or Fusion.
- Initially Download-Only: The first version of Cattery is download-only with no in-app browser or community submission portal yet, limiting discoverability and contribution workflows.
Frequently Asked Questions
What is Foundry Cattery?
Cattery is a free online library of machine learning models in .cat format that can be loaded directly into Foundry's Nuke compositing software via the Inference node, enabling artists to apply AI-powered effects without writing code.
What is a .cat file?
A .cat file is the model format used by Nuke's Inference node. Models trained with CopyCat or third-party PyTorch models that have been converted are stored in this format so Nuke can interpret and run them.
Do I need to know machine learning to use Cattery?
No. Cattery is designed so that artists without an ML or engineering background can simply download a model and load it into Nuke's Inference node to start applying effects immediately.
Which version of Nuke do I need to use Cattery?
You need Nuke 13 or later, which introduced the CopyCat and Inference node ML toolset. Nuke 13.1 and above also supports loading third-party .cat models.
Can I submit my own models to Cattery?
Foundry's long-term roadmap includes opening Cattery to user submissions, allowing artists to contribute models trained in CopyCat or converted from PyTorch. The initial release is download-only from the Foundry website.
