T2M-GPT: Generating Human Motion from Textual Descriptions with Discrete Representations
This work addresses text-to-motion generation for applications like animation or robotics, but it is incremental as it builds on existing VQ-VAE and GPT methods.
The authors tackled human motion generation from text descriptions using a VQ-VAE and GPT framework, achieving a lower FID score of 0.116 compared to 0.630 for a diffusion-based method on the HumanML3D dataset.
In this work, we investigate a simple and must-known conditional generative framework based on Vector Quantised-Variational AutoEncoder (VQ-VAE) and Generative Pre-trained Transformer (GPT) for human motion generation from textural descriptions. We show that a simple CNN-based VQ-VAE with commonly used training recipes (EMA and Code Reset) allows us to obtain high-quality discrete representations. For GPT, we incorporate a simple corruption strategy during the training to alleviate training-testing discrepancy. Despite its simplicity, our T2M-GPT shows better performance than competitive approaches, including recent diffusion-based approaches. For example, on HumanML3D, which is currently the largest dataset, we achieve comparable performance on the consistency between text and generated motion (R-Precision), but with FID 0.116 largely outperforming MotionDiffuse of 0.630. Additionally, we conduct analyses on HumanML3D and observe that the dataset size is a limitation of our approach. Our work suggests that VQ-VAE still remains a competitive approach for human motion generation.