Tencent’s flagship generative AI 3D modeling product that generates high-quality 3D models from text or images.
2026-01-28
Image Processing
Pricing:
$0.02/Point
starting from
Bulk order? Contact your manager for exclusive deals
稳定性
Stable
API Overview
Hunyuan 3D is Tencent’s flagship generative AI 3D modeling product, capable of generating high-quality 3D models from a single image or text prompt. Unlike previous 3D AIs that produced blurry, low-resolution meshes, Hunyuan adopts a two-stage Diffusion Transformer (DiT) architecture, separating geometry generation from texture synthesis, thereby enhancing creative efficiency and design flexibility, empowering multiple industries to efficiently produce 3D content. Its core positioning is as a “high-fidelity 3D content generation engine supporting multi-modal inputs,” specifically designed to provide end-to-end, efficient modeling solutions for scenarios including gaming, film and television, e-commerce, industrial design,3D printing, and home furnishings and furniture.Architectural Innovation: It employs a two-stage DiT architecture: Hunyuan 3D-DiT generates structurally coherent raw meshes, while Hunyuan 3D-Paint simultaneously projects multi-view PBR textures (albedo/normal/roughness/metallicity), achieving decoupled optimization of geometry and materials.
Version Matrix: The Professional Edition supports text-to-3D, image-to-3D, sketch-to-3D, eight-view input, and intelligent topology; the Speed Edition completes model generation in just 1 minute and 30 seconds, balancing speed with usability.
Comprehensive Functionality: It covers the entire 3D pipeline—from single-image/text generation, automatic UV unwrapping, component recognition, high-poly mesh reduction (Polygon 1.5), texture generation to output in STL/FBX/USDZ formats.
Leading Quality: It preserves microscopic structural details, with sharp edges and regular topologies, delivering high fidelity for complex objects such as furniture, characters, and products.
Scenario Implementation: It can be applied to real-world production workflows, including 3D printing (with editable models for complex structures), home decoration (automatically generating furniture with PBR materials), and e-commerce (single-image “shoot-and-model” functionality).
Generate 3D Models Based on HunYuan Large Model
Intelligently generate 3D models from input text descriptions/images.
3 concurrent tasks are provided by default, meaning a maximum of 3 submitted tasks can be processed at the same time; the next task can only start after the previous one is completed.
Different parameter combinations correspond to different deduction standards when calling the interface. The default mode is Text/Image-to-3D + Standard Texture Mode, which costs 0.4 PTC per time.
✅ No fees are deducted for failed generation tasks (including failures caused by content risk control policies, applicable to all failure reasons).
Generation Task Types GenerateType is a required parameter with a default value of Normal; you can select 1 of the 4 generation modes, and fees are deducted according to the corresponding type.
Generation Task Type
Function Description
Cost
Normal
Generate textured geometric models
0.4 PTC/time
LowPoly
Generate textured geometric models after intelligent decimation
Generate textured geometric models from input sketches/line drawings
0.5 PTC/time
Additional Parameters
Can be used additionally after selecting the generation task type; multiple parameters are supported, and fees are deducted by superposition according to the additional items.
Additional Parameter Name
Function Description
Additional Cost
MultiViewImages
Generate 3D models from multi-view images
0.2 PTC/time
EnablePBR
Generate 3D models with PBR materials, including metallicity, roughness, and normal maps
0.2 PTC/time
FaceCount
Generate 3D models with custom face counts
0.2 PTC/time
ResultFormat
Customize the format of the generated models
0.1 PTC/time
Key Parameter Rules
Required (select one of three): One of ImageBase64, ImageUrl, and Prompt must be filled in; Prompt cannot coexist with ImageBase64/ImageUrl (except for Sketch mode).
Version Restrictions: The LowPoly parameter is not supported in version 3.1; only version 3.1 supports top/bottom/oblique front views.
Parameter Validity: The EnablePBR parameter is invalid in Geometry mode; PolygonType is only valid in LowPoly mode.
Request Parameters
Header Parameters
AuthorizationstringRequired
Example Value:Bearer {{YOUR_API_KEY}}
Request Body application/json
Modelenum<string>Optional
🔖 Hybrid Generative 3D Generation Model Version
Default: 3.0
Note: LowPoly parameter is not available when version 3.1 is selected
Enum Value:
3.03.1
Example Value:3.0
PromptstringOptional
✨ Vincent 3D Chinese Forward Prompt Words
Character limit: up to 1024 utf-8 characters
Filling rules: Required with ImageBase64, ImageUrl three choices, and can not exist at the same time with the latter two
Example Value:一只小猫
ImageBase64stringOptional
🖼️ Tuscan 3D Base64 encoded image
Filling rules: required for Prompt and ImageUrl, and cannot exist at the same time as Prompt
Auxiliary instructions: Prioritize the generation of 3D models based on image encoding, corresponding to the needs of 3D content description
By default, 3 concurrent tasks are provided, which means that up to 3 submitted tasks can be processed at the same time; after the previous task is processed, the next task can be started.
Generate 3D Models Based on HunYuan Large Model
Intelligently generate 3D models from input text descriptions/images.
1 concurrent task is provided by default, meaning a maximum of 1 submitted task can be processed at the same time; the next task can only start after the previous one is completed.
Different parameter combinations correspond to different deduction standards when calling the interface. The default mode is Text/Image-to-3D + Standard Texture Mode, which costs 0.3 PTC per time.
Generation Parameters
Basic generation parameters with a unified base fee for both text-to-3D and image-to-3D generation, no additional differentiation.
Generation Parameter
Function Description
Cost
Prompt
Generate 3D models from text descriptions
0.3 PTC/time
ImageUrl/ImageBase64
Generate 3D models from images (URL/Base64)
0.3 PTC/time
Additional Parameters
Can be optionally used in addition to basic generation; parameters support separate activation, and fees are deducted by superposition according to the additional items.
Additional Parameter Name
Function Description
Additional Cost
EnablePBR
Generate 3D models with PBR materials, including metallicity, roughness and normal maps
0.2 PTC/time
Request Parameters
Header Parameters
AuthorizationstringRequired
Example Value:Bearer {{YOUR_API_KEY}}
Request Body application/json
PromptstringOptional
✨ Vincent 3D Core Parameters
Function: Chinese forward description prompt for 3D content
Limit: Up to 1024 utf-8 characters
-Rule: Required with ImageBase64, ImageUrl three choices, and can not exist at the same time with the latter two
ImageBase64stringOptional
🖼️ Tucson 3D (Base64 format)
Function: Generate 3D content based on the input image, Base64 encoded string of the incoming image
Limit: Up to 1024 utf-8 characters
Rules: Required with Prompt, ImageUrl, and cannot exist at the same time as Prompt
ImageUrlstringOptional
🔗 Tucson 3D (URL format)
Function: Generate 3D content based on the input image, valid network URL of the incoming image
Resolution limitation: image unilateral resolution ≥ 128 and ≤ 5000
Size limit: File size ≤ 8M
Supported formats: jpg, png, jpeg, webp
Rules: Required with Prompt, ImageBase64, and cannot exist at the same time as Prompt
ResultFormatenum<string>Optional
📦 3D model output format
Function: Specify a single output format for generating 3D models
Default rule: when not specified, close Geometry to return obj, glb; enable Geometry to return only glb
Optional format: the following enumerated values
Enum Value:
STLUSDZFBX
Example Value:STL
EnablePBRenum<boolean>Optional
🎨 PBR Material Generation Switch
Function: Whether to turn on Physics Based Rendering (PBR) material generation
Default: false (not enabled)
Enum Value:
truefalse
EnableGeometryenum<boolean>Optional
🧱 Single Geometry White Mode Generation Switch
Function: Whether to turn on single geometry generation, and generate a non-textured 3D white mold after turning on
Format restrictions: OBJ format is not supported after opening, default output GLB format
Chaotic Variant 3D Extreme Edition Result Acquisition Interface
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Hybrid generative 3D interface, after entering a single geometric model and a reference diagram or text description, the corresponding texture map can be generated.
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Default interface request frequency limit: 20 times/second.
Price: 0.6 PTC/time
Request Parameters
Header Parameters
AuthorizationstringRequired
Example Value:Bearer {{YOUR_API_KEY}}
Request Body application/json
File3DobjectRequired
📂 Source 3D model file (required)
Contains two required subfields: Model Format Type and Model File URL (Url)
Type only supports specified enumeration format, not customizable
Typeenum<string>Required
📦 3D model file format type
Enum Value:
OBJGLB
Example Value:GLB
UrlstringRequired
Valid network URL for 🔗 3D model file
PromptstringOptional
✨ Vincent 3D Chinese Forward Prompt Words
Character limit: up to 200 utf-8 characters
Filling rules: Required with Image, ImageUrl three choices, and can not exist at the same time with the latter two
ImagestringOptional
🖼️ 3D model texture reference diagram (Base64 encoded or web URL)
Priority rules: Choose between Base64 and Url, both of which give preference to Url
Resolution limitation: image unilateral resolution ≥ 128 and ≤ 4096
Size limit: Base64 encoded string size ≤ 10MB
Format restrictions: jpg, jpeg, png supported
EnablePBRenum<boolean>Optional
🎨 PBR Material Generation Switch
Feature Description: Generate a Physical Based Rendering (PBR) material after opening to enhance the authenticity of the 3D model texture
Default value: false
Enum Value:
truefalse
Hunyuansheng 3D texture generation (Get Task Result)
Hybrid Elemental 3D Texture Generation Result Acquisition Interface
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Default interface request frequency limit: 20 times/second.
The hybrid generative 3D interface adopts the Polygon 1.5 model. After inputting the 3D high mode, a 3D model with neat wiring and a lower surface number can be generated.
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Default interface request frequency limit: 20 times/second.
Price: 1 PTC/time
Request Parameters
Header Parameters
AuthorizationstringRequired
Example Value:Bearer {{YOUR_API_KEY}}
Request Body application/json
File3DobjectRequired
📂 Source 3D model file (required)
Contains two required subfields: Model Format Type, Model File Network URL (Url)
Type only supports enumeration format, cannot be customized
Typeenum<string>Required
📦 3D model file format type
Enum Value:
OBJGLB
Example Value:GLB
UrlstringRequired
Valid network URL for 🔗 3D model file
PolygonTypeenum<string>Optional
🔺 Model Polygon Mesh Type
Function Description: Defines the mesh composition shape of the 3D model surface
Default: triangle
Optional value description:
triangle
quadrilateral
Enum Value:
trianglequadrilateral
Example Value:triangle
FaceLevelenum<string>Optional
📏 Model Subtracted Numeric Gear Type
Function description: Define the number of faces retention gear (accuracy level) after the 3D model reduction operation
Optional value description:
high: high grade (keep a large number of faces, model details are more complete)
medium: mid-range (balancing details with file volume)
low: low grade (retains a small number of faces and has the smallest file size)
Enum Value:
highmediumlow
Example Value:high
Hunyuansheng 3D Intelligent Topology (Get Task Result)
Hybrid Genesis 3D Smart Topology Result Acquisition Interface
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Default interface request frequency limit: 20 times/second.
The hybrid-origin 3D interface is based on the hybrid-origin large model and intelligently generates 3D based on the input text description/picture.
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Default interface request frequency limit: 20 times/second.
Price: 0.6 PTC/time
Request Parameters
Header Parameters
AuthorizationstringRequired
Example Value:Bearer {{YOUR_API_KEY}}
Request Body application/json
ProfileobjectOptional
🖼️ Live avatar reference image (Base64 encoding/web URL supported)
Priority rules: Choose between Base64 and Url, both of which give preference to Url
Resolution limitation: image unilateral resolution ≥ 500 and ≤ 4096
Size limit: Base64 encoded string size ≤ 10MB
Format restrictions: jpg, jpeg, png supported
UrlstringRequired
🔗 Live Avatar Reference Image Web URL
Templateenum<string>Optional
🎭 3D Character Generation Template (select the corresponding style to generate characters)
Hybrid Elemental 3D Character Generation Result Acquisition Interface
By default, 1 concurrent task is provided, which means that up to 1 submitted task can be processed at the same time. After the previous task is processed, the next task can be started.
Default interface request frequency limit: 20 times/second.
Price: 0 PTC/time
Request Parameters
Path Parameters
JobIdstringRequired
Header Parameters
AuthorizationstringRequired
Example Value:Bearer {{YOUR_API_KEY}}
API Pricing
$￥ 円 ₽
Model
Description
302.AI Price
Hunyuansheng 3D Professional Edition
The corresponding deduction is made based on the selected parameters
$0.02/Point
Hunyuansheng 3D Professional Edition (Get Task Result)
Get Task Result
Free
Hunyuansheng 3D Extreme Edition
The corresponding deduction is made based on the selected parameters
$0.02/Point
Hunyuansheng 3D Extreme Edition (Get Task Result)
Get Task Result
Free
Hunyuansheng 3D texture generation
-
$0.6/call
Hunyuansheng 3D texture generation (Get Task Result)
Get Task Result
Free
Hunyuansheng 3D Intelligent Topology
-
$1/call
Hunyuansheng 3D Intelligent Topology (Get Task Result)
Get Task Result
Free
Hunyuansheng 3D UV unfolding
-
$0.2/call
Hunyuansheng 3D UV unfolding (Get Task Result)
Get Task Result
Free
Hunyuansheng 3D component generation
-
$0.6/call
Hunyuansheng 3D component generation (Get Task Result)
Get Task Result
Free
Hunyuansheng 3D character generation
-
$0.6/call
Hunyuansheng 3D character generation (Get Task Result)
