中文阅读

HY-Motion 1.0: Scaling Flow Matching Models for 3D Motion Generation

🔥 News

• Jan 29, 2026: 📊 We released the evaluation prompts and code for SSAE (Structured Semantic Alignment Evaluation), a VLM-based metric designed to assess the semantic alignment of generated videos. Check the ssae directory for usage details!
• Dec 30, 2025: 🤗 We released the inference code and pretrained models of HY-Motion 1.0. Please give it a try via our HuggingFace Space and our Official Site!

Introduction

HY-Motion 1.0 is a series of text-to-3D human motion generation models based on Diffusion Transformer (DiT) and Flow Matching. It allows developers to generate skeleton-based 3D character animations from simple text prompts, which can be directly integrated into various 3D animation pipelines. This model series is the first to scale DiT-based text-to-motion models to the billion-parameter level, achieving significant improvements in instruction-following capabilities and motion quality over existing open-source models.

Key Features

• State-of-the-Art Performance: Achieves state-of-the-art performance in both instruction-following capability and generated motion quality.

• Billion-Scale Models: We are the first to successfully scale DiT-based models to the billion-parameter level for text-to-motion generation. This results in superior instruction understanding and following capabilities, outperforming comparable open-source models.

• Advanced Three-Stage Training: Our models are trained using a comprehensive three-stage process:

  • Large-Scale Pre-training: Trained on over 3,000 hours of diverse motion data to learn a broad motion prior.

  • High-Quality Fine-tuning: Fine-tuned on 400 hours of curated, high-quality 3D motion data to enhance motion detail and smoothness.

  • Reinforcement Learning: Utilizes Reinforcement Learning from human feedback and reward models to further refine instruction-following and motion naturalness.

🎁 Model Zoo

HY-Motion 1.0 Series

Model	Description	Date	Size	Huggingface	VRAM (min)
HY-Motion-1.0	Standard Text2Motion Model	2025-12-30	1.0B	Download	26GB
HY-Motion-1.0-Lite	Lightweight Text2Motion Model	2025-12-30	0.46B	Download	24GB

Note: To reduce GPU VRAM requirements, please use the following settings: --num_seeds=1, text prompt with less than 30 words, and motion length less than 5 seconds.
Note: This table does not includes GPU VRAM requirements for LLM-based prompt engineering feature. If you have sufficient VRAM to run HY-Motion-1.0 model but gradio fails with a VRAM-related error, Run the Gradio application with prompt engineering disabled by setting the environment variable like this: DISABLE_PROMPT_ENGINEERING=True python3 gradio_app.py

🤗 Get Started with HY-Motion 1.0

HY-Motion 1.0 supports macOS, Windows, and Linux.

• Code Usage (CLI)
• Gradio App

1. Installation

First, install PyTorch via the official site. Then install the dependencies:

git clone https://github.com/Tencent-Hunyuan/HY-Motion-1.0.git
cd HY-Motion-1.0/
# Make sure git-lfs is installed
git lfs pull
pip install -r requirements.txt

2. Download Model Weights

Please follow the instructions in ckpts/README.md to download the necessary model weights.

Code Usage (CLI)

We provide a script for local batch inference, suitable for processing large amounts of prompts.

# HY-Motion-1.0
python3 local_infer.py --model_path ckpts/tencent/HY-Motion-1.0

# HY-Motion-1.0-Lite
python3 local_infer.py --model_path ckpts/tencent/HY-Motion-1.0-Lite

Common Parameters:

• --input_text_dir: Directory containing .txt or .json prompt files.
• --output_dir: Directory to save results (default: output/local_infer).
• --disable_duration_est: Disable LLM-based duration estimation.
• --disable_rewrite: Disable LLM-based prompt rewriting.
• --prompt_engineering_host / --prompt_engineering_model_path: (Optional) Host address / local checkpoint for the Duration Prediction & Prompt Rewrite Module.
  • Download: You can download the Duration Prediction & Prompt Rewrite Module from Here.
  • Note: If you do not set these parameter, you must also set --disable_duration_est and --disable_rewrite. Otherwise, the script will raise an error due to host unavailable.

Gradio App

You can host a Gradio web interface on your local machine for interactive visualization:

python3 gradio_app.py

After running the command, open your browser and visit http://localhost:7860

Prompting Guide & Best Practices

1. Language & Length: Please use English. For optimal results, keep your prompt under 60 words. For other languages, please use the Text2MotionPrompter to rewrite the prompt.

2. Content Focus: Focus on action descriptions or detailed movements of the limbs and torso.

3. Current Limitations (NOT Supported):

• ❌ Non-humanoid Characters: Animations for animals or non-human creatures.
• ❌ Subjective/Visual Attributes: Descriptions of complex emotions, clothing, or physical appearance.
• ❌ Environment & Camera: Descriptions of objects, scenes, or camera angles.
• ❌ Multi-person Interactions: Motions involving two or more people.
• ❌ Special Modes: Seamless loop or in-place animations.

4. Example Prompts:

• A person performs a squat, then pushes a barbell overhead using the power from standing up.
• A person climbs upward, moving up the slope.
• A person stands up from the chair, then stretches their arms.
• A person walks unsteadily, then slowly sits down.

🔗 BibTeX

If you found this repository helpful, please cite our reports:

@article{hymotion2025,
  title={HY-Motion 1.0: Scaling Flow Matching Models for Text-To-Motion Generation},
  author={Tencent Hunyuan 3D Digital Human Team},
  journal={arXiv preprint arXiv:2512.23464},
  year={2025}
}

🤗 Community Integrations

We appreciate the community for creating integrations for HY-Motion! Here are some third-party implementations:

• ComfyUI-HY-Motion1 by @jtydhr88

Acknowledgements

We would like to thank the contributors to the FLUX, diffusers, HuggingFace, SMPL/SMPLH, CLIP, Qwen3, PyTorch3D, kornia, transforms3d, FBX-SDK, GVHMR, and HunyuanVideo repositories or tools, for their open research and exploration.