Using the Generate Image API with Image5

This page explains the features and workflows available with the Image5 model.

Overview

The Image5 model is a new model that is available with the Generate Image API. It is a more advanced model that is designed to generate more realistic images.

This endpoint supports two operations, Text‑to‑Image generation and Image‑to‑Image instruct edit, through the same request payload. The mode is determined by content in the referenceBlobs field.

About Text‑to‑Image generation

Text‑to‑Image generation is performed when referenceBlobs is empty → The API generates a new image using the prompt. Use this mode to create entirely new image content from a text prompt.

Example Request Payload

{
  "prompt": "A hyper-detailed illustration of a floating city at sunrise",
  "aspectRatio": "16:9",
  "modelId": "firefly_image",
  "numVariations": 1,
  "referenceBlobs": []  // Empty array indicates Text‑to‑Image generation
}

About Image‑to‑Image instruct edit

Image‑to‑Image instruct edit is performed when referenceBlobs contains one or more images → The API performs an instruct edit on the provided image(s). Use this mode to apply prompt‑driven edits to an existing image while preserving its structure and composition.

Example Request Payload

{
  "prompt": "Add warm sunset lighting and enhance the reflection on the water",
  "modelId": "firefly_image",
  "numVariations": 1,
  "referenceBlobs": [
    {
      "source": {
        "uploadId": "123e4567-e89b-12d3-a456-426614174000" // Upload ID of the refernce image(s)
      },
      "usage": "general"
    }
  ]
}