BAMM: Human Motion Modeling & Architecture
January 2026
25 min read
Generative Models, Motion Models, Transformers, BAMM
Use arrow keys or click to navigate slides. Press 'F' or Fullscreen icon for best experience.
What You'll Learn
- •The Challenge of Current Text-to-Motion Models
- •BAMM Architecture: Motion Tokenizer & Masked SA Transformer
- •Training with Hybrid Attention Masking
- •Inference using Cascaded Motion Decoding
- •Classifier-Free Guidance (CFG)
Key Concepts Covered
Strategy combining masking patterns for robust training.
Iterative decoding process for finer motion control.
Discretizing continuous motion data.
Resources
Slide Overview
- Challenges & Architecture (Slides 1-10)
- Training Methodologies (Slides 11-20)
- Inference & Guidance (Slides 21-28)
- Evaluation & Results (Slides 29-end)