Skip to content

README.md

Latest commit

 

History

History
218 lines (153 loc) · 5.65 KB

README.md

File metadata and controls

218 lines (153 loc) · 5.65 KB

Hunyuan 3D

Tencent Hunyuan AI 3D Generation (OpenAI-compatible API) - Text-to-3D and image-to-3D generation.

⚠️ Important

This skill uses OpenAI-compatible API, different from traditional Tencent Cloud API:

  • Uses API Key authentication (not SecretId/SecretKey)
  • OpenAI-style interface
  • Professional version only (no rapid version)

Features

  • 🎨 Text-to-3D: Generate 3D models from text descriptions
  • 🖼️ Image-to-3D: Generate 3D models from images
  • 🎯 Multi-version: Support model versions 3.0 and 3.1

Quick Start

1. Enable Hunyuan 3D Service

Required: Enable service in Tencent Cloud console first:

  1. Visit Tencent Hunyuan 3D Console
  2. Click "Enable Now" or "Apply"
  3. Read and agree to service agreement
  4. Wait for service activation (usually instant)

Common Issues:

  • If "Insufficient Resources" shows, service is initializing, wait 5-10 minutes
  • If cannot enable, may need real-name verification or contact support

2. Get API Key

  1. Visit Hunyuan 3D API Key Management
  2. Click "Create API Key"
  3. Enter name (e.g., hunyuan-3d-key)
  4. Copy the generated API Key (format: sk-xxxxx)

⚠️ Important: API Key is shown only once, please save it properly!

Alternative: If above link doesn't work, use Hunyuan API Key page

3. Configure Environment Variable

Windows PowerShell:

# Temporary (current session)
$env:HUNYUAN_3D_API_KEY = "sk-xxxxx"

# Permanent (recommended)
[Environment]::SetEnvironmentVariable("HUNYUAN_3D_API_KEY", "sk-xxxxx", "User")

Linux/Mac:

# Temporary
export HUNYUAN_3D_API_KEY="sk-xxxxx"

# Permanent (add to ~/.bashrc or ~/.zshrc)
echo 'export HUNYUAN_3D_API_KEY="sk-xxxxx"' >> ~/.bashrc
source ~/.bashrc

4. Verify Installation

# Check environment variable
Write-Host "API Key: $($env:HUNYUAN_3D_API_KEY.Substring(0,15))..."

# Test generation
python scripts/generate.py --mode text --prompt "a cute dog"

If "Insufficient Resources" error: Service is initializing, wait 5-10 minutes and retry

Usage

Basic Usage

# Text-to-3D
python scripts/generate.py --mode text --prompt "a cute pig"

# Image-to-3D
python scripts/generate.py --mode image --image-url "https://example.com/pig.jpg"

# Use version 3.1
python scripts/generate.py --mode text --prompt "dog" --model 3.1

Parameters

Parameter Description Example
--mode Generation mode text (text-to-3D), image (image-to-3D)
--prompt Text description (text-to-3D) "a cute pig"
--image-url Image URL (image-to-3D) "https://example.com/pig.jpg"
--model Model version 3.0 (default), 3.1
--output Output directory ./models

Model Versions

Version Features
3.0 Default, full features
3.1 New version, but LowPoly and Sketch parameters not available

Input Requirements

Text-to-3D:

  • Text description: max 1024 UTF-8 characters
  • Chinese prompts supported

Image-to-3D:

  • Format: jpg, png, jpeg, webp
  • Size: ≤8MB
  • Resolution: 128px ~ 5000px (single side)

Output

Generated 3D models are saved to {output}/{date}/{job_id}/:

  • model.glb - 3D model file (GLB format)
  • model.obj - 3D model file (OBJ format, in ZIP)
  • preview.png - Preview image
  • info.json - Task information

Output Formats:

  • Both GLB and OBJ formats returned
  • Includes texture maps
  • With preview image

API Information

  • Base URL: https://api.ai3d.cloud.tencent.com
  • Submit Task: POST /v1/ai3d/submit
  • Query Task: POST /v1/ai3d/query
  • Authentication: API Key (Header: Authorization: sk-xxxxx)

Troubleshooting

Service Not Enabled

Error: ResourceUnavailable.NotExist

Solution:

  1. Visit https://console.cloud.tencent.com/ai3d
  2. Enable Hunyuan 3D service
  3. Wait a few minutes for activation
  4. Retry

Insufficient Resources

Error: ResourceInsufficient

Solution:

  • Service just enabled, wait 5-10 minutes and retry
  • Or contact Tencent Cloud support

API Key Error

Error: Unauthorized or authentication failed

Solution:

Response Format

OpenAI-compatible API returns:

{
  "Response": {
    "JobId": "xxx",
    "Status": "DONE",
    "ResultFile3Ds": [...]
  }
}

Note: Status field is Status not StatusCode, success value is DONE not SUCCESS

Important Notes

  1. Async API: Divided into submit task and query task steps
  2. Task Validity: 24 hours
  3. Concurrent Limit: Default 3 concurrent
  4. Professional Only: OpenAI-compatible API doesn't support rapid version
  5. Generation Time: 3D generation takes longer (1-5 minutes), please be patient

Examples

# Generate pig 3D model
python scripts/generate.py --mode text --prompt "a cute pig, pink, cartoon style"

# Image-to-3D
python scripts/generate.py --mode image --image-url "https://example.com/pig-photo.jpg"

# Use version 3.1
python scripts/generate.py --mode text --prompt "dog" --model 3.1

Links

License

MIT License

中文版本