WuyinKeji GPT-Image-2

Generate images using the WuyinKeji GPT-Image-2 API (速创科技). Supports text-to-image, reference images, multiple aspect ratios, and async result polling. Use w...

WuyinKeji GPT-Image-2 Image Generation

This skill wraps the WuyinKeji (速创科技) GPT-Image-2 async image generation API.

API Endpoints

Action	Method	URL
Submit task	POST	https://api.wuyinkeji.com/api/async/image_gpt
Query result	GET	https://api.wuyinkeji.com/api/async/detail?id=<task_id>&key=<key>

Authentication

• Header: Authorization: <api_key>
• Body: also include "key": "<api_key>" in JSON payload

1. Submit Generation Task

Request

curl -s -X POST "https://api.wuyinkeji.com/api/async/image_gpt" \
  -H "Content-Type: application/json" \
  -H "Authorization: <API_KEY>" \
  -d '{
    "key": "<API_KEY>",
    "prompt": "Your image generation prompt here",
    "size": "16:9",
    "urls": ["http://example.com/ref.jpg"]
  }'

Parameters

Field	Required	Type	Description
key	Yes	string	API key
prompt	Yes	string	Generation prompt
size	No	string	Aspect ratio: auto, 16:9, 9:16, 1:1, 3:2, 2:3 (default: auto)
urls	No	array	Reference image URLs (up to 14 images supported by the model)

Response

{
  "code": 200,
  "msg": "成功",
  "data": {
    "id": "image_xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
    "count": "1"
  }
}

Save the data.id — this is the async task ID.

2. Query Result (Polling)

curl -s "https://api.wuyinkeji.com/api/async/detail?id=<TASK_ID>&key=<API_KEY>"

Response Fields

Field	Meaning
status: 0	Task submitted, waiting to start
status: 1	Generating in progress
status: 2	Done — check data.result[0] for image URL

Example (completed)

{
  "code": 200,
  "data": {
    "status": 2,
    "result": ["https://openpt1.wuyinkeji.com/xxxx.png"],
    "created_at": "2026-04-22 13:42:01",
    "updated_at": "2026-04-22 13:43:17"
  }
}

3. Download Image

curl -sL "<RESULT_URL>" -o output.png

Typical resolution: 1672×941 (the API output is fixed at this resolution regardless of the size parameter; use the size parameter mainly for aspect ratio hints).

Model Capabilities

• Text rendering: Good Chinese/English text support in images (prompt explicitly for sharp, readable typography)
• Reference images: Supports up to 14 reference URLs for style/person consistency
• Realism: Improved skin texture, hair, facial details over earlier DALL·E models
• Color accuracy: Reduced yellow/orange color cast compared to DALL·E

Tips for Quality

1. For sharp text: Explicitly prompt "sharp, crisp, perfectly legible Chinese typography" and specify font style
2. For person consistency: Include a clear reference photo URL in urls
3. For billboard/posters: Prompt "billboard quality", "high contrast", "professional advertising photography"
4. Output resolution is fixed at ~1672×941 — upscaling must be done client-side if 4K is needed

Helper Script

A Bash helper script is available at scripts/generate.sh:

./scripts/generate.sh \
  -k "<API_KEY>" \
  -p "A beautiful sunset over mountains" \
  -s "16:9" \
  -r "http://example.com/ref.jpg" \
  -o "output.png"

The script handles submit → poll → download in one command.