Hunyan3D-2mv - A multi-view 3D asset generation model by Tencent, part of the Hunyan3D 2.0 project.

## Purpose of Hunyan3D-2mv **Hunyan3D-2mv** is designed to generate **high-resolution textured 3D assets** from multiple 2D input views (e.g., front, left, and back views). It is optimized for applications like game development, virtual reality, and metaverse content creation, enabling rapid and realistic 3D modeling. ## Comparison with Hunyan3D-2mini - **Hunyan3D-2mv**: Focuses on **multi-view shape control** and high-resolution output (1.1B parameters). - **Hunyan3D-2mini**: A **lightweight variant** (0.6B parameters) optimized for low-configuration devices, sacrificing some detail for faster inference and lower resource usage. ## Technical Features - **Multi-view input**: Supports front/left/back views for precise shape alignment. - **High-resolution textures**: Generates detailed geometry and vibrant textures via the integrated Hunyan3D-Paint model. - **Smart mesh decimation**: Reduces polygon count for performance optimization. - **Open-source**: Code and models are freely accessible on [GitHub](https://github.com/Tencent/Hunyan3D-2) and [Hugging Face](https://huggingface.co/tencent/Hunyan3D-2mv). - **Accelerated framework**: Supports CUDA for faster generation. ## Usage Instructions Developers can integrate Hunyan3D-2mv via the `hy3dgen` library. Example code: ### Citation sources: - [Hunyan3D-2mv](https://huggingface.co/tencent/Hunyan3D-2mv) - Official URL Updated: 2025-04-01