Private AI

Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Search
 
 
 
 
All conversations

Create

Chat
Studio

Explore

Models
Pricing
AI Apps
Blog

Library

Gallery
Workspace
Usage & billing
API
Suggestions and bugs
Use cases

|

Terms

|

Privacy

|

Refunds
Usage & billing
Teams
Referrals
Presets
Settings
FAQPress & mediaBrand assetsBenchmarksNano RPCEarn crypto
  1. Models
  2. 3D

Hunyuan3D V3

Generate

Best-quality textured 3D from an image or prompt: film-quality geometry with baked color textures, optional PBR materials and low-poly output. About two minutes; files can reach 30 MB.

Added Sep 8, 2026

Details

Price per model
$0.63/model
Model type
both
Typical duration
~2 min
Inputs
Image + Text

Price table

Each row is the price for one completed 3D model generation.

VariantPrice

Geometry

$0.45/model

Geometry + PBR

$0.63/model

Normal

Default

$0.63/model

Normal + PBR

$0.81/model

LowPoly

$0.72/model

LowPoly + PBR

$0.9/model

Settings

Generation controls available for this model.

Output Format

GLB

Face count

Number

Default

500000

Target polygon count for the mesh (40,000 to 1,500,000).

Generation type

Select

Default

Normal

Options (3)

Normal (textured), Low poly, Geometry only

Normal produces a textured mesh. LowPoly optimizes for games. Geometry skips textures and is cheapest.

PBR materials

Toggle

Default

No

Adds physically based material maps. Costs extra for every generation type, including Geometry.

Polygon type

Select

Default

triangle

Options (2)

Triangles, Quads

Triangle meshes import everywhere; quads suit sculpting and subdivision workflows.

API request

Submit the generation with POST /api/generate-video, then poll the returned run ID for completion.

{
  "model": "wavespeed-ai/hunyuan3d-v3",
  "prompt": "A studio-quality 3D model of a ceramic desk lamp.",
  "generate_type": "Normal",
  "enable_pbr": false,
  "polygon_type": "triangle",
  "face_count": 500000
}
Browse the 3D models API catalog

Examples

Loading examples…

Related 3d models

Compare Hunyuan3D V3 with similar models from the same provider or model family.

Hunyuan 3D 3.1 Rapid

wavespeed-ai/hunyuan-3d-v3.1-rapid

Fast, low-cost draft mesh from a single image or a short prompt (up to 200 characters). Geometry only, no textures, in under a minute. Use Hunyuan3D V3 or Tripo for a textured result.

TRELLIS.2

wavespeed-ai/trellis-2/image-to-3d

Turn one object image into a textured GLB with UVs and embedded PBR materials. Choose geometry and texture detail, control the target face count, and optionally preserve an existing transparent background.

Meshy 6

wavespeed-ai/meshy6

Meshy 6 turns a prompt (up to 600 characters) or up to four views of an object into a textured or untextured 3D mesh. Control topology, polygon count and texture resolution, with optional character rigging and animation. Generations typically take several minutes.

Meshy T2

meshy/t2

Meshy T2 creates 3D models with Smart Topology and separate triangle-mesh parts from a text prompt or one JPG/PNG image up to 20 MB. T2 is the faster variant. Choose 100–15,000 target triangles (default 4,000), geometry only or textures, optional PBR materials, humanoid rigging and animation. Download GLB and available OBJ, FBX or USDZ outputs. Text prompts support up to 600 characters; texture prompts up to 800. N1 texturing uses reference imagery and produces 2K base color without PBR.

Meshy T2.5

meshy/t2.5

Meshy T2.5 creates 3D models with Smart Topology and separate triangle-mesh parts from a text prompt or one JPG/PNG image up to 20 MB. Choose 100–30,000 target triangles (default 15,000), geometry only or textures, optional PBR materials, humanoid rigging and animation. Download GLB and available OBJ, FBX or USDZ outputs. Text prompts support up to 600 characters; texture prompts up to 800. N1 texturing uses reference imagery and produces 2K base color without PBR.

MoGe 2

microsoft/moge-2

MoGe 2 by Microsoft reconstructs visible scene geometry from one JPG, PNG or WebP image up to 2048 × 2048 pixels and 20 MB. Download a textured GLB mesh, a colored PLY point cloud, colorized depth and validity maps, and camera field of view and intrinsics. Choose one of four checkpoints; normal-capable checkpoints also return surface normals. Processing detail, masking, edge filtering and an optional known horizontal field of view are adjustable. Reconstruction does not recover unseen surfaces or guarantee absolute scale. $0.01 per started decimal megapixel, minimum one megapixel; all options included.