MeshGen enables users to control Blender with natural language prompts, leveraging AI agents for 3D modeling tasks. It targets Blender users seeking to integrate AI assistance into their workflow, offering a flexible architecture with multiple inference backends for both local and remote execution.

How It Works

MeshGen acts as a Blender addon that translates natural language commands into actions within the 3D modeling software. It supports various inference backends, including local execution via llama.cpp or Ollama, and remote API access through Hugging Face, Anthropic, or OpenAI. This modular design allows users to choose between cost-free local processing (requiring significant GPU resources) or leveraging powerful cloud-based models. Optional integrations with LLaMA-Mesh and Hyper3D extend its capabilities for local mesh understanding/generation and high-fidelity mesh creation, respectively.

Quick Start & Requirements

• Installation: Download the addon ZIP from the latest release and install via Blender's Add-on preferences.
• Local Backend: Requires a powerful NVIDIA GPU with at least 8GB VRAM and a CUDA-enabled build. A recommended Llama 3.1 8B GGUF model can be downloaded directly.
• Remote Backends: Ollama (local server), Hugging Face, Anthropic, or OpenAI APIs. API keys are required for remote services.
• Optional Integrations: LLaMA-Mesh also requires an NVIDIA GPU with 8GB VRAM and a remote API backend. Hyper3D requires an API key.
• Documentation: Usage is initiated via the N-panel or Sidebar in Blender. Troubleshooting involves checking the system console.

Highlighted Details

• Supports local inference with llama.cpp and Ollama for cost-free operation.
• Integrates with Hugging Face, Anthropic, and OpenAI APIs for broad model access.
• Optional LLaMA-Mesh integration for local mesh understanding and generation.
• Optional Hyper3D integration for high-fidelity 3D mesh generation.

Maintenance & Community

The project is hosted on GitHub. Users can report errors via the Issues tab.

Licensing & Compatibility

The license is not explicitly stated in the provided README.

Limitations & Caveats

Local inference and LLaMA-Mesh integration have significant hardware requirements (NVIDIA GPU with 8GB+ VRAM). Hyper3D mesh generation may take several minutes per mesh.