What is HY-Motion 1.0
HY-Motion 1.0 (Hunyuan Motion 1.0) is a billion-parameter-scale large model for text-to-3D motion generation open-sourced by the Tencent Hunyuan team. Based on the Diffusion Transformer (DiT) architecture and flow matching mechanism, it can generate high-fidelity, smooth, and diverse 3D character skeletal animations from just a single natural language description. The model adopts a full-stage training strategy, including pre-training on over 3,000 hours of diverse motion data, fine-tuning on 400 hours of curated high-quality data, and optimization through reinforcement learning combined with human feedback and a reward model. It covers more than 200 motion types across 6 major categories, outputs in SMPL-H skeletal format, and supports atomic motions, composite sequences, and concurrent motion generation. In performance evaluations, its instruction-following capability reaches 78.6%, and its average motion quality score is 3.43 (on a 5-point scale), comprehensively surpassing open-source baselines such as MoMask and DART.
Main Features of HY-Motion 1.0
-
Text-Driven Motion Generation: Directly generates high-quality 3D skeletal animations from natural language descriptions, enabling efficient transformation from text to motion.
-
Diverse Motion Coverage: Covers more than 200 motion types across 6 major categories, including basic locomotion, sports and athletics, social and leisure, and game character motions, meeting the needs of various scenarios.
-
High-Quality Motion Output: Supports SMPL-H skeletal format, and the generated animations are smooth and natural with rich motion details, suitable for demanding animation production.
-
Compatibility with Mainstream Tools: Directly compatible with mainstream 3D tools such as Blender, Unity, and Unreal Engine, making it convenient for users to quickly integrate and use.
-
Flexible Output Options: Supports atomic motions, composite sequences, and concurrent motion generation, meeting animation needs of different complexity levels.
-
Open Source and Ease of Use: Provides complete inference code, pre-trained model weights, and detailed documentation, supports multiple operating systems, and is easy to get started with.
Technical Principles of HY-Motion 1.0
-
Based on Diffusion Transformer Architecture: Adopts the advanced Diffusion Transformer (DiT) architecture, combined with a flow matching mechanism, to efficiently transform text descriptions into high-quality 3D skeletal animations.
-
Full-Stage Training Strategy: The model optimizes performance through three stages of training: pre-training on over 3,000 hours of diverse motion data to learn broad motion patterns; fine-tuning on 400 hours of curated high-quality data to improve motion precision; and finally optimization through reinforcement learning combined with human feedback and a reward model to ensure the naturalness and accuracy of generated motions.
-
Flow Matching Mechanism: By leveraging flow matching technology, the model can better capture the continuity and dynamic changes in motion data, generating smoother and more natural animations.
-
Multimodal Fusion: Deeply fuses the text modality with the motion modality, converting text descriptions into feature vectors through a text encoder, which are then combined with the motion generator to achieve precise motion generation.
-
Reinforcement Learning Optimization: Introduces a reinforcement learning mechanism, combined with human feedback and a reward model, to evaluate and optimize generated motions, further improving the model's motion generation quality and instruction-following capability.
Project Address of HY-Motion 1.0
- Project Website：https://hunyuan.tencent.com/motion
- Github Repository：https://github.com/Tencent-Hunyuan/HY-Motion-1.0
- Huggingface Model Library：https://huggingface.co/tencent/HY-Motion-1.0
- arXiv Technical Paper：https://arxiv.org/pdf/2512.23464
Application Scenarios of HY-Motion 1.0
-
Film and Animation Production: Quickly generate character motions, reduce animation production costs and time, and improve creative efficiency.
-
Game Development: Provide diverse motions for game characters, enhance the gaming experience, and adapt to different game scenarios.
-
Virtual Streamers and Digital Humans: Generate natural and smooth motions to enhance the interactivity and realism of virtual avatars.
-
Education and Training: Create instructional animations, simulate complex motion processes, and assist teaching and training.
-
Advertising and Marketing: Produce personalized animated advertisements, attract audience attention, and improve marketing effectiveness.
-
VR/AR Applications: Generate real-time motions for virtual reality and augmented reality content to enhance immersion.