Hunyuan3D Multiview

Hunyuan3D-2mv is the technical designation for Tencent's Hunyuan3D v2 Multiview model. It is a multi-view image-to-3D generation system that uses front, back, and left profile images to eliminate backside ambiguity and reconstruct complete, watertight 3D meshes.

The standard Multi-view endpoint runs the full 2.6B parameter flow-based diffusion transformer (Hunyuan3D-DiT) without distillation, producing maximum geometric edge crispness and fidelity for 23.5 base credits. The Turbo endpoint uses step distillation to accelerate generation down to sub-30 second turnaround times at 20.7 base credits, with minor softening on micro-details.

Choose the Turbo variant during early drafting, interactive workflows, and rapid asset exploration when turnaround speed matters most. Switch to standard Multi-view when preparing final production assets that demand maximum silhouette precision and clean surface contours.

Enabling the textured setting triggers the 1.3B parameter Hunyuan3D-Paint model, which bakes illumination-invariant PBR texture maps directly onto the mesh. This increases generation compute cost by 3x. Leaving texturing disabled outputs a clean, untextured matte white mesh ideal for custom hand-painting or shader authoring in external DCC tools.

The model accepts three isolated orthographic or turnaround views: Front, Back, and Left. For optimal mesh synthesis, inputs should feature isolated subjects against transparent or solid white backgrounds with flat, neutral lighting, matching scale, and vertical alignment across all three perspectives.

Hunyuan3D Multiview excels at low-to-mid poly game props, stylized characters, organic creatures, and e-commerce objects. It is not designed for hollowed interior cavities, transparent or refractive glass surfaces, fine strand-based hair, or high-tolerance mechanical CAD parts.