Hunyuan3D 2 Turbo

Hunyuan3D 2 Turbo is a high-speed 3D asset generation model developed by Tencent, representing the accelerated evolution of the Hunyuan3D 2.0 system. Designed for rapid production workflows, it utilizes the FlashVDM (Video Diffusion Model) acceleration framework to reduce inference times significantly, enabling the generation of high-quality 3D meshes from single images or text prompts in under ten seconds. This version is approximately 30 times faster than the standard generation pipeline.

The model follows a specialized two-stage architecture. The first stage employs Hunyuan3D-DiT, a flow-based diffusion transformer with approximately 1.1 billion parameters, which synthesizes the core geometric structure (the bare mesh). The second stage, Hunyuan3D-Paint, applies high-resolution, multi-view consistent textures to the generated mesh. This decoupled approach allows the system to maintain high geometric fidelity while producing production-ready PBR (Physically Based Rendering) materials including diffuse, normal, metallic, and roughness maps.

Hunyuan3D 2 Turbo is optimized for both professional 3D artists and general creators, supporting various input formats and offering integration with industry tools like Blender. For optimal results in image-to-3D tasks, users should provide clear, centered source images with high contrast. When utilizing text-to-3D features, descriptive prompts that specify the object's form, structural details, and material surface properties yield the most accurate and consistent spatial reconstructions.