Image Generation

Generate images with any model available on ImageRouter.

Choose your API endpoint

Endpoint	Form-Data encoded endpoint	JSON encoded endpoint
Text-to-Image	✅	✅
Image-to-Image	✅	❌
Compatibility	GPT Image (edit)	GPT Image

If you plan to generate images from both text and image, this kind of mixed usage is supported by the Form-Data encoded endpoint. This simplifies the integration to just one endpoint for both Text-to-Image and Image-to-Image generation. I’d only use the JSON endpoint, if Form-Data is not an option.

Both APIs are compatible with OpenAI GPT-Image API, but some OpenAI parameters may be ignored by some models.

Form-Data endpoint

Supports both Text-to-Image and Image-to-Image generation. Requests are encoded as multipart/form-data.

curl 'https://api.imagerouter.io/v1/openai/images/edits' \
-H 'Authorization: Bearer YOUR_API_KEY' \
-F 'prompt=YOUR_PROMPT' \
-F 'model=openai/gpt-image-1' \
-F 'quality=auto' \
-F 'size=auto' \
-F 'response_format=url' \
-F 'output_format=webp' \
-F 'image[]=@your_image1.webp' \
-F 'image[]=@your_image2.webp' \
-F 'mask[]=@your_mask.webp'

const formData = new FormData()
formData.append('prompt', 'YOUR_PROMPT')
formData.append('model', 'openai/gpt-image-1')
formData.append('quality', 'auto')
formData.append('size', 'auto')
formData.append('response_format', 'url')
formData.append('output_format', 'webp')

// Add your image files (up to 16)
const imageFile1 = await fetch('your_image1.webp').then(r => r.blob())
formData.append('image[]', imageFile1)

// or from local file
const imagePath = '/home/phoenics/projects/Image router/materials/logo.png'
const imageBuffer = fs.readFileSync(imagePath)
const imageBlob = new Blob([imageBuffer], { type: 'image/png' })
formData.append('image[]', imageBlob, path.basename(imagePath))

// Add mask file - some models support/require it
//const maskFile = await fetch('your_mask.webp').then(r => r.blob())
//formData.append('mask[]', maskFile)

const response = await fetch('https://api.imagerouter.io/v1/openai/images/edits', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY'
  },
  body: formData
})

const data = await response.json()
console.log(data)

import requests

url = "https://api.imagerouter.io/v1/openai/images/edits"
headers = {
  "Authorization": "Bearer YOUR_API_KEY"
}

payload = {
  "prompt": "YOUR_PROMPT",
  "model": "openai/gpt-image-1",
  "quality": "auto",
  "size": "auto",
  "response_format": "url",
  "output_format": "webp"
}

with open("your_image1.webp", "rb") as img1, \
   open("your_image2.webp", "rb") as img2, \
   open("your_mask.webp", "rb") as mask:
  files = {
      "image[]": [img1, img2],
      "mask[]": mask
  }
  response = requests.post(url, files=files, data=payload, headers=headers)
  print(response.json())

note: /v1/openai/images/generations and /v1/openai/images/edits are the exact same, available at 2 URLs for compatibility reasons.

Parameters

model required Image model to use for generation.
prompt optional Text input for generating images. Many models require prompt, but not all.
quality optional Models supporting this parameter have “quality” feature label here
- auto [default] - Usually points to “medium”, but some models (eg. gpt-image-1) automatically adjust quality based on your prompt.
- low
- medium
- high
size optional Accepted values are different for each model. Some models and providers completely ignore size.
- auto [default] - Uses the default recommended size for each model. Some models (eg. gpt-image-1) automatically adjust size based on your prompt.
- WIDTHxHEIGHT (eg 1024x1024)
response_format optional
- url [default] - Returns the image URL hosted on ImageRouter’s servers. The image is saved in your logs. and is publicly accessible with the URL if you share it.
- b64_json - Returns the image as base64-encoded JSON data. The image is saved in your logs. and is publicly accessible with the URL if you share it.
- b64_ephemeral - Same as b64_json but the image is not saved in our system. The provider may still have it.
output_format optional Image format for the generated output.
- webp [default]
- jpeg
- png
image[] optional Input file for Image-to-Image generation. (Supported models have “edit” label here.)
mask[] optional Image editing mask (most models don’t need this).

Response

The OpenAI compatible response makes sure you can use ImageRouter anywhere GPT-Image is already implemented.

{
  "created": 1769286389027, // timestamp
  "data":[
    {
      "url": "https://storage.imagerouter.io/fffb4426-efbd-4bcc-87d5-47e6936bf0bb.webp"
      // or "b64_json": "...", if you select a different response_format
    }
  ],
  "latency": 6942,
  "cost": 0.004
}

JSON endpoint

This endpoint exists for compatibility reasons (so ImageRouter can be used with the OpenAI SDK). File uploads are not possible with application/json encoding, so Image-to-Image generation is not supported for this endpoint.

curl 'https://api.imagerouter.io/v1/openai/images/generations' \
-H 'Authorization: Bearer YOUR_API_KEY' \
--json '{
  "prompt": "YOUR_PROMPT",
  "model": "test/test",
  "quality": "auto",
  "size": "auto",
  "response_format": "url",
  "output_format": "webp"
}'

const url = 'https://api.imagerouter.io/v1/openai/images/generations'
const payload = {
  prompt: 'YOUR_PROMPT',
  model: 'test/test',
  quality: 'auto',
  size: 'auto',
  response_format: 'url',
  output_format: 'webp'
}

const response = await fetch(url, {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify(payload)
})

const data = await response.json()
console.log(data)

import requests

url = "https://api.imagerouter.io/v1/openai/images/generations"
payload = {
  "prompt": "YOUR_PROMPT",
  "model": "test/test",
  "quality": "auto",
  "size": "auto",
  "response_format": "url",
  "output_format": "webp"
}
headers = {
  "Authorization": "Bearer YOUR_API_KEY"
}

response = requests.post(url, json=payload, headers=headers)
print(response.json())

Please contact me if you miss anything.