Hunyuan 3D 3.1 Pro Image to 3D

Hunyuan 3D 3.1 Pro Image to 3D — Image-to-3D Model

What is Hunyuan 3D 3.1 Pro Image to 3D?

Hunyuan 3D 3.1 Pro is Tencent's professional image-to-3D model. It turns a single front image — or up to eight reference views — into a production-ready, textured 3D asset you can export as GLB or OBJ. Built on a two-stage pipeline that generates geometry first and then synthesizes texture, version 3.1 adds improved multi-view diffusion and sparse-view reconstruction, giving cleaner topology, watertight meshes, and fewer artifacts at joints and thin structures. Drop the output straight into Unity, Unreal Engine, Blender, or a WebXR scene.

Key Features

• •Single-image or multi-view (up to eight angles) image-to-3D reconstruction.
• •Normal mode returns a textured model; Geometry mode returns a bare, untextured white mesh.
• •Optional PBR materials — metallic, roughness, and normal maps that engines read natively.
• •Configurable target face count from 40,000 to 1,500,000 polygons.
• •Clean quad topology, watertight meshes, and smarter UV unwrapping.
• •Standard GLB and OBJ output compatible with Unity, Unreal, and Blender.

Best Use Cases

Game studios use Hunyuan 3D 3.1 Pro to prototype props, weapons, and cosmetics from concept art in minutes. E-commerce teams convert product photos into rotatable 360-degree viewers. Design and visualization teams build reviewable asset libraries, and makers export watertight meshes for 3D printing. In testing, a single carved-rooster photo reconstructed into a fully textured GLB with its silhouette, relief, and paint intact; a steampunk hand drill with PBR enabled showed correct metallic response; and four consistent views of a tin robot produced a coherent, all-around model.

Prompt Tips and Output Quality

This is an image-conditioned model, so input quality drives output quality. Use a single, well-lit subject centered on a plain background, filling over half the frame, with minimal occlusion. Front-facing or isometric views reconstruct best. Add back and side views for asymmetric objects to resolve geometry the model would otherwise guess. Enable PBR for hard-surface assets bound for a game engine, raise face count for fine relief, and switch to Geometry mode when you plan to retopologize or texture downstream.

How to Use Hunyuan 3D 3.1 Pro Image to 3D

Hunyuan 3D 3.1 Pro converts a product or character image into a textured, engine-ready 3D model. Follow these steps and settings to get clean, usable results on the first run.

How it works

Upload a front image of one object on a plain background. The model builds the geometry, then paints it, and returns a GLB (and OBJ) you can preview in the viewer and download. Add more views, or toggle materials and polygon budget, to match your pipeline.

Recommended Settings

• •input_image_url: a single, centered subject on a plain background, well-lit and filling most of the frame. JPG, PNG, or WEBP, 128-5000px, under 8MB.
• •generate_type: Normal for a finished textured asset; Geometry for a bare mesh to retopologize or texture later.
• •enable_pbr: true for hard-surface props that need metallic, roughness, and normal response under engine lighting; leave off for quick textured previews.
• •face_count: 500,000 is a solid default. Raise toward 1,500,000 for fine relief and hero assets; drop toward 40,000 for lightweight previews.
• •Multi-view: add back, left, and right (and more) for asymmetric objects to improve all-around accuracy.

Example inputs to try

Each example below ran end to end at the settings shown; the silhouette, surface relief, and texture of the input carried into the mesh.

• •Hero — single image to textured 3D (craft and handmade). Input: a hand-carved, painted folk-art rooster on a plain background. The output was a fully textured GLB with the red comb and painted tail intact — keep generate_type Normal and one centered subject for this.
• •Optional PBR materials (industrial). Input: a brass-and-steel steampunk hand drill. With enable_pbr true, the brass gears and steel read as real metal under light — use it for game-engine props and expect a small surcharge.
• •Geometry mode (sci-fi and fantasy). Input: a coiled winged dragon statuette. generate_type Geometry returned a clean untextured mesh preserving every scale — use it when you will texture or retopologize downstream.

Reference-guided (multi-view)

Hunyuan 3D 3.1 Pro takes up to eight views. The multi-view example used four: the first (front) image is the main subject, and the back, left, and right images refine the sides and rear. Supply them at the same scale and lighting with the subject identical across angles — four views of a wind-up tin robot reconstructed into a coherent, all-around model with its chest dial and riveted panels intact.

Further inputs the model is built for: 'a game-ready treasure chest with iron bands', 'a hard-surface sci-fi helmet', 'a ceramic figurine with fine relief'.

What inputs does it accept?

One front image, or up to eight views — front, back, left, right, top, bottom, left-front, and right-front.

What formats does it output?

Textured GLB and OBJ in Normal mode, or a bare white mesh in Geometry mode.

Does it support PBR materials?

Yes — metallic, roughness, and normal maps, optionally, in Normal mode.

How detailed can the mesh be?

Up to 1,500,000 faces, controlled by the face count setting.

Can I use it for games and 3D printing?

Yes; output is watertight with clean topology and standard formats for Unity, Unreal, Blender, and slicers.