This ComfyUI plugin provides text-to-3D human motion generation capabilities, leveraging the HY-Motion 1.0 model. It targets users seeking to create 3D animations from textual descriptions, offering features like automatic prompt optimization and flexible export options, thereby streamlining the animation pipeline.

How It Works

The core functionality is built upon the HY-Motion 1.0 model for generating 3D human motion from text. It integrates Large Language Models (LLMs), specifically Qwen3-8B, to automatically rewrite and optimize input text prompts and estimate motion duration, enhancing the quality and control of the generated motion. The system supports GGUF quantized models for reduced VRAM consumption and offers CPU offload for LLM components, making it more accessible on systems with limited GPU memory.

Quick Start & Requirements

• Installation: Navigate to ComfyUI/custom_nodes/ComfyUI-HY-Motion1 and run pip install -r requirements.txt.
• Model Weights: Download HY-Motion-1.0 (or Lite), clip-vit-large-patch14, and optional LLM models (Qwen3 variants, GGUF) into the ComfyUI/models/HY-Motion/ckpts/ directory.
• Prerequisites: Python, ComfyUI environment. Optional: auto-awq for AWQ models, fbxsdkpy for FBX export.
• VRAM: HY-Motion-1.0 requires ~8GB+, Lite ~4GB+. Text encoder (Qwen3-8B) VRAM varies significantly with quantization (e.g., AWQ ~3-4GB, bnb-4bit ~4-5GB, int8 ~8GB, full ~16GB). GGUF Q4_K_M needs ~5GB. Prompter LLM requires ~2-3GB. CPU offload is available for LLMs.
• Documentation: Node documentation is detailed within the README.

Highlighted Details

• Text-to-Motion Generation: Creates 3D human motion directly from text prompts.
• Prompt Optimization: Utilizes LLMs for automatic prompt rewriting and motion duration estimation.
• Previews: Features real-time skeleton preview and an interactive Three.js 3D animation viewer with playback controls.
• Export Options: Supports direct GLB export (no dependencies) and FBX export, including custom character retargeting (Mixamo supported).
• Memory Efficiency: Offers GGUF model support and CPU offload for LLMs to reduce VRAM requirements.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or project roadmap are provided in the README.

Licensing & Compatibility

The license refers to the original HY-Motion 1.0 project. No explicit compatibility notes for commercial use or closed-source linking are mentioned.

Limitations & Caveats

FBX export necessitates the installation of fbxsdkpy, which may present installation challenges; the GLB export is a dependency-free alternative. GGUF models require manual download and placement into the specified directory.