Seedream 5 Pro · Gensta API Docs

Text to image

POST/v1/generationsmodel: seedream-5.0-pro · mode: image_generation

json

{
  "model": "seedream-5.0-pro",
  "mode": "image_generation",
  "variant": "normal",
  "parameters": {
    "size": "1.5K",
    "num_outputs": 1,
    "prompt_optimization": "standard",
    "output_format": "jpeg",
    "watermark": false
  },
  "input": {
    "prompt": "A glass perfume bottle on dark stone, studio light",
    "images": []
  }
}

num_outputs means alternative outputs. Seedream 5 Pro currently accepts and returns exactly 1; repeating a request is a separate billable generation.

Image references and editing

Pass from zero to ten images. With images present, describe the desired transformation in prompt. Each item can be a stable HTTPS URL or a ready media_asset_id.

json

{
  "model": "seedream-5.0-pro",
  "mode": "image_generation",
  "variant": "normal",
  "parameters": {
    "size": "2K",
    "num_outputs": 1,
    "prompt_optimization": "fast",
    "output_format": "png",
    "watermark": false
  },
  "input": {
    "prompt": "Keep the product shape, place it in a snowy forest",
    "images": [
      { "type": "https_url", "url": "https://cdn.example.com/product.png" },
      { "type": "media_asset", "media_asset_id": "7c74c900-8b68-4b0f-a337-32ed3a9a2020" }
    ]
  }
}

Caller-hosted HTTPS images may be JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC, or HEIF. Reusable uploads accept JPEG, PNG, or WebP. Each image is limited to 30,000,000 bytes and 36,000,000 pixels, and each side must be at least 15 pixels.

For point- or box-guided edits, include the marker in the input image and refer to it in the prompt. This uses the same image_generation mode.

Parameters

Field	Accepted values	Meaning
size	1K, 1.5K, 2K, or WIDTHxHEIGHT	Output size; custom dimensions must satisfy the limits returned by GET /v1/models
num_outputs	1	Number of alternative outputs, not number of layers
prompt_optimization	standard or fast	Prompt optimization mode
output_format	png or jpeg	Output image format
watermark	boolean	Whether to add a generated-content watermark

Pricing in USD

GET /v1/models returns the current Gensta price for an image up to 2.61 megapixels, above 2.61 megapixels, and the additional charge for each input image after the first. The create response returns billing.reserved_amount; after success, billing.charged_amount contains the final USD charge.