ML for the Houdini Artist | SideFX | Gnomon HIVE 2025

Channel:

Houdini

Subscribers:

138,000

Published on April 3, 2025 7:24:33 PM ● Video Link: https://www.youtube.com/watch?v=6UrNFXRbP3s

Duration: 0:00

4,084 views

101

n this talk, technical artist Josh Karlin from SideFX provides a beginner-friendly overview of machine learning in Houdini, explaining core concepts like data collection, model training, and inference. He breaks down the muscle deformer example to demonstrate how to use Houdini's ML nodes, covering essential terms like ONNX and PCA, and discusses practical applications for character effects and simulations, all while emphasizing the ability to train models with minimal coding.

00:00 - Introduction & Overview
01:13 - What is Machine Learning?
02:22 - Step 1: Data Collection (Most Important)
03:17 - Step 2: Model Training
03:36 - Step 3: Inference (Making Predictions)
04:36 - Key Machine Learning Terms
08:34 - Reverse Engineering the Muscle Deformer Example
09:44 - Analyzing the Final Result (Inference)
11:51 - Creating the Onyx Model (Training)
13:46 - Understanding the Input Data (ML Example Output)
15:52 - Deep Dive into the ML Example Node
17:09 - Creating the Muscle Deformer Data Set
20:01 - Optimizing the Data Set (PCA)
22:50 - Finalizing the Data Set and Training
24:29 - Conclusion & Further Learning
25:01 - Q&A: Other Use Cases (Character Effects)
26:15 - Q&A: Pyro and Fluid Sim Possibilities
27:18 - Q&A: Onyx and Unreal/Maya Integration
28:50 - Q&A: Data Requirements and Synthetic Data
31:10 - Q&A: Onyx File Size and Optimization
33:23 - Q&A: Defining Output Data (Name Tags)
35:53 - Q&A: Serializing Input Data
37:05 - Q&A: Symmetry in Data Generation
39:41 - Q&A: Downsides of ML Muscle Deformation
41:40 - Q&A: Limitations of Onyx Inference Output
43:40 - Closing Remarks


#Houdini #MachineLearning #ML #VFX #3D #TechnicalArt #HoudiniHIVE #Gnomon