Images

​

Generate

result = client.images.generate(
    model="openai/gpt-image-2",
    prompt="A paper boat drifting on a rainy street, cinematic lighting",
    quality="medium",
    extra_body={"generation_config": {"resolution": "2K", "aspect_ratio": "16:9"}},
)
print(result.data[0].url)

curl https://api.geekidea.ai/v1/images/generations \
  -H "Authorization: Bearer $GEEKIDEA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-image-2",
    "prompt": "A paper boat drifting on a rainy street, cinematic lighting",
    "quality": "medium",
    "generation_config": { "resolution": "2K", "aspect_ratio": "16:9" }
  }'

​

Response

{
  "id": "img_01a119547bc77689a77ca1ae7686af8d",
  "created": 1791425100,
  "data": [
    {
      "url": "https://…",
      "width": 1402,
      "height": 1122,
      "previews": { "preview": "https://…", "lite": "https://…", "grid": "https://…", "icon": "https://…" }
    }
  ],
  "usage": { "input_tokens": 10, "output_tokens": 200, "total_tokens": 210, "images": 1, "cost": 0.00608, "currency": "USD",
             "cost_details": { "input": 0.00008, "cached_input": 0, "output": 0.006 } }
}

• url is a signed link to the original image, valid for at least a day. Download what you need to keep. To get fresh links later, call GET /v1/generations/{id} with the id.
• previews are WebP thumbnails in four sizes.
• b64_json replaces url when you set response_format: "b64_json".

​

Edit with reference images

result = client.images.edit(
    model="google/nano-banana-pro",
    image=[open("product.png", "rb"), open("style.jpg", "rb")],
    prompt="Place the product on a marble table in this style",
)

{
  "model": "google/nano-banana-pro",
  "prompt": "Place the product on a marble table in this style",
  "images": [
    { "image_url": "https://example.com/product.png" },
    { "image_url": "data:image/png;base64,iVBORw0..." },
    { "file_id": "file-0192f1c3a7b87c3e9a4f5b6c7d8e9f01" }
  ]
}

​

How images are priced

• Per image: a price per image, optionally by quality or resolution tier (image_per_spec). The tier is matched on generation_config.resolution first, then quality; tiers not listed use image.
• Per token (GPT Image, FLUX.2, MAI Image): input tokens × input + output image tokens × output. Higher resolution and quality produce more output tokens. A low-quality 1K GPT Image uses about 200 output tokens.

​

Midjourney

​

Long-running requests