HY-Motion 1.0

From Tencent

Billion-Parameter Open-Source Text-to-3D Human Motion Generation – Scaling Flow Matching for Realistic Skeleton Animations

Pricing Model

Starting Price

$0/Month

Visit HY-Motion 1.0
Last Updated: January 5, 2026
By Zelili AI

About This AI

HY-Motion 1.0 is a series of state-of-the-art text-to-3D human motion generation models from Tencent’s Hunyuan team, released December 30, 2025.

Built on Diffusion Transformer (DiT) architecture with Flow Matching training, it scales to 1 billion parameters (standard version) and 460 million (Lite variant), producing high-quality skeleton-based 3D character animations from natural language prompts.

The model excels in instruction-following, motion diversity, temporal coherence, and physical realism across 200+ motion categories including locomotion, sports actions, dance, gestures, and complex multi-step behaviors.

Outputs are in standard formats (SMPL-H skeletons) compatible with Blender, Unity, Unreal Engine, and other 3D pipelines for direct import and use.

Key strengths include significant improvements over prior open-source models in semantic alignment, motion quality, and artifact reduction (e.g., foot sliding) via large-scale data, fine-tuning, and physics-based rewards.

Fully open-source under Tencent Hunyuan Community license with inference code, pretrained weights, Hugging Face Space demo, and official site for easy testing.

Supports text prompts in English/Chinese, zero-shot generalization, and integration into animation workflows without motion capture or manual keyframing.

Ideal for game developers, animators, VFX artists, researchers, and creators needing fast, high-fidelity 3D motion assets from simple descriptions.

Key Features

  1. Text-to-3D Motion Generation: Converts natural language prompts into realistic skeleton-based 3D human animations
  2. Billion-Parameter Scale: 1B standard model for superior quality and instruction following; 460M Lite for efficiency
  3. Flow Matching DiT Architecture: Combines diffusion transformers with flow matching for coherent, diverse motions
  4. 200+ Motion Categories: Covers locomotion, sports, dance, gestures, complex actions, and multi-step behaviors
  5. High Semantic Alignment: Strong prompt adherence with minimal artifacts like foot sliding or unnatural poses
  6. Standard Output Formats: SMPL-H skeletons exportable to FBX, BVH, GLB for Blender, Unity, Unreal Engine
  7. Zero-Shot Generalization: Generates novel motions without specific training examples
  8. Hugging Face Space Demo: Interactive online testing with prompt input and animation preview
  9. Open-Source Full Stack: Inference code, pretrained weights, and evaluation tools on Hugging Face/GitHub
  10. Multilingual Prompt Support: Works well with English and Chinese text descriptions

Price Plans

  1. Free ($0): Fully open-source with pretrained models, inference code, Hugging Face demo, and no usage restrictions under Tencent Hunyuan Community license
  2. Enterprise/Cloud (Custom): Potential future hosted options or premium support via Tencent (not available yet)

Pros

  1. State-of-the-art open-source performance: Outperforms prior models in motion quality, diversity, and prompt adherence
  2. Completely free and open: No usage fees; Apache-compatible license with full code and weights
  3. Easy integration: Direct export to popular 3D engines and pipelines
  4. High scalability: Billion-parameter power enables complex, realistic animations
  5. Online demo available: Test via Hugging Face Space without local setup
  6. Physics-aware training: Reduces common artifacts for more natural results
  7. Rapid release momentum: Inference code and models available immediately on announcement

Cons

  1. Requires GPU for local inference: Heavy model needs strong hardware for reasonable speed
  2. Recent release: Community support, fine-tuning examples, and integrations still emerging
  3. No hosted API pricing: Fully local/self-hosted; no official cloud service mentioned
  4. Limited to skeleton motion: Outputs SMPL-H poses, not full textured/rigged characters
  5. Prompt sensitivity: Complex or ambiguous descriptions may require rewriting for best results
  6. No mobile/web native: Demo is HF Space; core use is developer/local deployment
  7. Evaluation ongoing: Human evals strong but broader benchmarks evolving

Use Cases

  1. Game animation prototyping: Generate character actions for Unity/Unreal without motion capture
  2. Film and VFX pre-vis: Create realistic motion sequences from script descriptions
  3. 3D content creation: Quickly produce animations for avatars, metaverse, or AR/VR
  4. Research in motion generation: Benchmark, fine-tune, or extend for new domains
  5. Educational tools: Visualize human movements for sports science or dance training
  6. AI agent simulation: Create training data for embodied AI with diverse motions
  7. Automated asset pipeline: Script batch generation for large-scale projects

Target Audience

  1. Game developers: Needing fast, high-quality character animations
  2. 3D animators and VFX artists: Accelerating pre-production workflows
  3. AI researchers: Studying large-scale motion models and flow matching
  4. Metaverse/AR/VR creators: Building dynamic avatar behaviors
  5. Developers using Unity/Unreal/Blender: Integrating text-driven motion
  6. Open-source enthusiasts: Experimenting with billion-parameter generative models

How To Use

  1. Visit Hugging Face Space: Go to huggingface.co/spaces/tencent/HY-Motion-1.0 for online demo
  2. Enter prompt: Type descriptive text (English/Chinese) like 'a person performing a karate kick'
  3. Generate animation: Click run to produce 3D motion sequence preview
  4. Download output: Export skeleton animation in supported formats
  5. Local setup: Clone GitHub repo Tencent-Hunyuan/HY-Motion-1.0
  6. Install dependencies: Follow README for PyTorch, transformers, etc.
  7. Run inference: Use provided scripts with downloaded weights for batch or custom use

How we rated HY-Motion 1.0

  • Performance: 4.6/5
  • Accuracy: 4.7/5
  • Features: 4.8/5
  • Cost-Efficiency: 5.0/5
  • Ease of Use: 4.3/5
  • Customization: 4.7/5
  • Data Privacy: 5.0/5
  • Support: 4.4/5
  • Integration: 4.6/5
  • Overall Score: 4.7/5

HY-Motion 1.0 integration with other tools

  1. Hugging Face: Model weights, inference code, and interactive demo space for testing
  2. GitHub Repository: Full open-source code and community contributions via Tencent-Hunyuan/HY-Motion-1.0
  3. 3D Engines: Direct export to Blender, Unity, Unreal Engine via FBX/BVH/GLB formats
  4. Animation Pipelines: Compatible with Maya, Houdini, or custom tools for motion integration
  5. Local Development: Runs on PyTorch with GPU acceleration; no external cloud required

Best prompts optimised for HY-Motion 1.0

  1. A professional basketball player performing a powerful slam dunk on a court, dynamic jump, ball in hand, crowd in background, realistic motion
  2. A graceful ballerina executing a perfect pirouette in a spotlight on stage, elegant dress flowing, smooth rotation and balance
  3. A ninja warrior performing a series of high-speed flips and sword strikes in a misty bamboo forest at night
  4. A person running energetically through a park in autumn, leaves falling, natural arm swing and stride
  5. A fitness trainer demonstrating a proper deadlift technique in a gym, slow controlled movement with perfect form
HY-Motion 1.0 marks a major advance in open-source text-to-3D motion generation, scaling to 1B parameters for superior quality, diversity, and prompt adherence across 200+ categories. Fully free with easy Hugging Face demo and engine compatibility, it’s excellent for game devs, animators, and researchers. Setup is developer-oriented, but results rival proprietary tools in realism and usability.

FAQs

  • What is HY-Motion 1.0?

    HY-Motion 1.0 is Tencent’s open-source text-to-3D human motion generation model series, scaling to 1B parameters with Diffusion Transformer and Flow Matching for realistic skeleton animations from text prompts.

  • When was HY-Motion 1.0 released?

    It was open-sourced on December 30, 2025, with inference code, pretrained models, Hugging Face Space demo, and official site made available.

  • Is HY-Motion 1.0 free to use?

    Yes, it is completely free and open-source with full weights and code on Hugging Face under Tencent Hunyuan Community license; no usage fees.

  • What are the model sizes in HY-Motion 1.0?

    Standard version has 1 billion parameters; Lite version has 460 million for lighter deployment while maintaining strong performance.

  • What formats does HY-Motion 1.0 output?

    It generates SMPL-H skeleton animations exportable in FBX, BVH, GLB for direct use in Blender, Unity, Unreal Engine, and other 3D tools.

  • How many motion categories does HY-Motion 1.0 support?

    It covers over 200 categories including locomotion, sports, dance, gestures, and complex multi-step actions with high diversity.

  • Where can I try HY-Motion 1.0 online?

    Use the interactive demo on Hugging Face Spaces at huggingface.co/spaces/tencent/HY-Motion-1.0; enter text prompts to generate animations.

  • Who developed HY-Motion 1.0?

    Developed by Tencent’s Hunyuan 3D Digital Human Team as part of their AI research and open-source efforts.

Newly Added Tools​

Qwen-Image-2.0

$0/Month

Qodo AI

$0/Month

Codiga

$10/Month

Tabnine

$59/Month
HY-Motion 1.0 Alternatives

Seedance 2.0

$0/Month

VideoGen

$12/Month

WUI.AI

$10/Month

About Author

Hi Guys! We are a group of ML Engineers by profession with years of experience exploring and building AI tools, LLMs, and generative technologies. We analyze new tools not just as a user, but as someone who understands their technical depth and real-world value.We know how overwhelming these tools can be for most people, that’s why we break down complex AI concepts into simple, practical insights. Our goal is to help you discover these magical AI tools that actually save your time and make everyday work smarter, not harder.“We don’t just write about AI: We build, test and simplify it for you.”