PackDiT: Joint Human Motion and Text Generation via Mutual Prompting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Zhongyu, Chai, Wenhao, Zhou, Zhuoran, Yang, Cheng-Yen, Huang, Hsiang-Wei, Hwang, Jenq-Neng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910802313740288
author Jiang, Zhongyu
Chai, Wenhao
Zhou, Zhuoran
Yang, Cheng-Yen
Huang, Hsiang-Wei
Hwang, Jenq-Neng
author_facet Jiang, Zhongyu
Chai, Wenhao
Zhou, Zhuoran
Yang, Cheng-Yen
Huang, Hsiang-Wei
Hwang, Jenq-Neng
contents Human motion generation has advanced markedly with the advent of diffusion models. Most recent studies have concentrated on generating motion sequences based on text prompts, commonly referred to as text-to-motion generation. However, the bidirectional generation of motion and text, enabling tasks such as motion-to-text alongside text-to-motion, has been largely unexplored. This capability is essential for aligning diverse modalities and supports unconditional generation. In this paper, we introduce PackDiT, the first diffusion-based generative model capable of performing various tasks simultaneously, including motion generation, motion prediction, text generation, text-to-motion, motion-to-text, and joint motion-text generation. Our core innovation leverages mutual blocks to integrate multiple diffusion transformers (DiTs) across different modalities seamlessly. We train PackDiT on the HumanML3D dataset, achieving state-of-the-art text-to-motion performance with an FID score of 0.106, along with superior results in motion prediction and in-between tasks. Our experiments further demonstrate that diffusion models are effective for motion-to-text generation, achieving performance comparable to that of autoregressive models.
format Preprint
id arxiv_https___arxiv_org_abs_2501_16551
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PackDiT: Joint Human Motion and Text Generation via Mutual Prompting
Jiang, Zhongyu
Chai, Wenhao
Zhou, Zhuoran
Yang, Cheng-Yen
Huang, Hsiang-Wei
Hwang, Jenq-Neng
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Human motion generation has advanced markedly with the advent of diffusion models. Most recent studies have concentrated on generating motion sequences based on text prompts, commonly referred to as text-to-motion generation. However, the bidirectional generation of motion and text, enabling tasks such as motion-to-text alongside text-to-motion, has been largely unexplored. This capability is essential for aligning diverse modalities and supports unconditional generation. In this paper, we introduce PackDiT, the first diffusion-based generative model capable of performing various tasks simultaneously, including motion generation, motion prediction, text generation, text-to-motion, motion-to-text, and joint motion-text generation. Our core innovation leverages mutual blocks to integrate multiple diffusion transformers (DiTs) across different modalities seamlessly. We train PackDiT on the HumanML3D dataset, achieving state-of-the-art text-to-motion performance with an FID score of 0.106, along with superior results in motion prediction and in-between tasks. Our experiments further demonstrate that diffusion models are effective for motion-to-text generation, achieving performance comparable to that of autoregressive models.
title PackDiT: Joint Human Motion and Text Generation via Mutual Prompting
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2501.16551