Hunyuan3D V3

byTencent AI Lab

Generate production-ready quad-dominant 3D meshes and full PBR materials directly from text, reference images, or rough sketches

Generate 3D Assets
Hunyuan3D V3

How Hunyuan3D V3 works

From initial concept prompt or reference sketch to a clean, textured GLB asset ready for your 3D viewport.

Provide Text or Visual Reference

Provide Text or Visual Reference

Enter a descriptive prompt, upload a reference image, or submit a line sketch to establish asset silhouette and materials.

Configure Topology and Material Maps

Configure Topology and Material Maps

Select polygon layout, adjust target face counts between 40k and 1.5M, and toggle PBR generation for complete texture channels.

Export Production-Ready GLB

Export Production-Ready GLB

Download the textured quad-dominant 3D mesh directly into your game engine, 3D DCC suite, or rendering pipeline.

What Hunyuan3D V3 is good at

A hierarchical 10B diffusion transformer engineered for clean quad topology, granular density controls, and physically based rendering maps.

Multimodal Generation Modes

Multimodal Generation Modes

Generate 3D assets from text descriptions, turn reference photographs into matching models, or use rough 2D sketch silhouettes to lock proportions while text guides materials.

Tunable Polygon Density

Tunable Polygon Density

Scale mesh density from 40,000 to 1,500,000 polygons to match your target platform, with quadrilateral retopology for sculpting or triangle layouts for direct game-engine import.

Full PBR Material Synthesis

Full PBR Material Synthesis

Synthesize complete Physically Based Rendering material sets including Albedo, Normal, Roughness, and Metallic channels for realistic light interaction in downstream engines.

Production-Ready Topology Workflows

Production-Ready Topology Workflows

Output clean untextured white meshes in Geometry mode for manual Substance Painter texturing, or generate streamlined LowPoly assets for real-time performance.

Made with Hunyuan3D V3

A curated selection of game-ready props, industrial mockups, and historical artifacts generated with rich surface materials.

Retro-futuristic espresso machine with brass and enamel finishes

Retro-futuristic espresso machine with brass and enamel finishes

Articulated medieval steel gauntlet with brass rivets

Articulated medieval steel gauntlet with brass rivets

High-performance ski goggles with lava-orange iridium lens

High-performance ski goggles with lava-orange iridium lens

Faceted crystal fragrance bottle with brushed gold hardware

Faceted crystal fragrance bottle with brushed gold hardware

Riveted iron mining lantern with glowing crystalline core

Riveted iron mining lantern with glowing crystalline core

What people build with Hunyuan3D V3

Built for 3D generalists, indie game developers, and industrial concept artists seeking fast, production-viable asset turnaround.

Game Asset Production

01

Produce game-ready background assets, weapons, and environmental props with automated quad-dominant retopology and PBR textures.

Industrial Concept Lookdev

02

Iterate on industrial product shapes, consumer electronics, and vehicle silhouettes using sketch-to-3D and text-guided lookdev.

E-Commerce 3D Visualizer

03

Convert 2D catalogue photographs into interactive 3D product previews and WebGL assets with accurate material properties.

Virtual Museum Artifacts

04

Reconstruct archaeological specimens, antique relics, and cultural artifacts from museum reference imagery into high-density meshes.

Spatial & XR Prototyping

05

Generate lightweight furniture, decorative items, and architectural set dressings directly for AR, VR, and spatial environments.

Frequently Asked Questions

Image-to-3D delivers the highest geometric fidelity by converting a 2D reference photo directly into a matching 3D mesh, while Text-to-3D generates novel assets from scratch based purely on descriptive prompts. Sketch-to-3D bridges the two by using 2D line art or silhouettes to lock in proportions and outlines while relying on the text prompt to synthesize materials, colors, and surface details.

Choose LowPoly mode when targeting real-time game engines like Unreal Engine or Unity to produce optimized quadrilateral or triangle meshes between 40k and 80k faces. If you plan to hand-paint or texture the mesh in external software like Substance Painter, select Geometry mode to output a clean, untextured white model at reduced cost.

Hunyuan3D V3 excels at generating static game props, hard-surface objects, environmental scenery, and e-commerce product mockups with PBR material channels. It is not designed for articulated character rigs, animation armatures, or precision CAD engineering tolerances, and it struggles to render fine readable typography on surfaces.

The face count parameter controls target polygon density between 40,000 and 1,500,000 faces. Use 40k to 80k faces for lightweight mobile and XR applications, 200k to 500k faces for standard desktop game props, and scale up to 1,000,000+ faces when generating high-detail assets for cinematic rendering or digital sculpting.

Hunyuan3D V3 was developed by Tencent AI Lab using a 10-billion parameter hierarchical 3D Diffusion Transformer (3D-DiT) architecture. It separates generation into a two-stage process: Hunyuan3D-DiT builds flow-based geometry across up to 3.6 billion voxels, while Hunyuan3D-Paint bakes high-resolution PBR textures.

Try Hunyuan3D V3 on Fuser

Generate production-ready quad-dominant 3D meshes and full PBR materials directly from text, reference images, or rough sketches

Generate 3D Assets