GPT Image 2.6 Sunburst Image Edit

curl --request POST \
  --url https://api.vidgo.ai/api/generate/submit \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "openai/gpt-image-2.6-sunburst/edit",
  "input": {
    "prompt": "Preserve the product and replace the background with a warm modern studio.",
    "image_urls": [
      "https://cdn.vidgo.ai/temp/reference-product.png"
    ],
    "resolution": "2K"
  }
}
'

{
  "code": 200,
  "data": {
    "task_id": "task-example-001",
    "status": "not_started",
    "created_time": "2026-10-09T10:30:00"
  }
}

{
  "detail": "prompt is required"
}

{
  "detail": "prompt is required"
}

GPT Image 2.6 Sunburst Image Edit edits images from text instructions: send 1–16 reference images to restyle, replace backgrounds, combine products into a new scene, or change only a masked region. Sunburst is the GPT Image 2.6 option for detailed visual work and precision-focused editing. For prompt-only generation, use GPT Image 2.6 Sunburst Text to Image (openai/gpt-image-2.6-sunburst/text-to-image). GPT Image 2.6 accepts the same parameters as GPT Image 2.5 Sunburst, plus output_compression, moderation and complexity.

​

Models

• openai/gpt-image-2.6-sunburst/edit — Image edit with GPT Image 2.6 Sunburst, 1–16 reference images, one image per request.

​

Features

• Instruction-based editing with up to 16 reference images, referenced by order in the prompt.
• Optional mask to limit the edit to one region.
• Three resolution tiers (1K, 2K, 4K) with aspect-ratio presets, auto, or custom WIDTHxHEIGHT sizes.
• Five quality tiers: low, medium, high, xhigh, max.
• Transparent or opaque backgrounds; PNG, JPEG or WebP output.
• Adjustable JPEG/WebP compression, moderation strictness and planning effort (complexity).

​

Required parameters

• prompt (string): Text instructions describing the edit. Up to 32,000 characters; must not be blank.
• image_urls (string[]): 1–16 reference images as publicly accessible HTTP(S) URLs, in the order you refer to them in the prompt.

​

Optional parameters

• size (string): Aspect-ratio preset, auto, or custom WIDTHxHEIGHT. When omitted, the output follows the first reference image’s aspect ratio. See Size.
• resolution (string): 1K, 2K or 4K. Default 1K.
• quality (string): low, medium, high, xhigh or max. Default high.
• background (string): auto, transparent or opaque. Default auto. A transparent background requires png or webp output.
• output_format (string): png, jpeg or webp. Default png.
• output_compression (integer): Compression level from 0 to 100. Only allowed when output_format is jpeg or webp. When omitted, the default compression applies.
• mask_url (string): Mask image selecting the area to edit, as a publicly accessible HTTP(S) URL. Requires image_urls.
• moderation (string): Content moderation strictness, auto or low. Default auto; low applies less restrictive filtering.
• complexity (string): Planning effort the model spends before generating, low, medium or high. Default high. Does not change the price.

​

Size

• Presets: 1:1, 2:3, 3:2, 4:3, 3:4, 4:5, 5:4, 16:9, 9:16, 21:9. When size is omitted, the output follows the first reference image’s aspect ratio, using the closest supported ratio at the selected resolution. Send size to override it.
• Presets keep the selected resolution tier; actual pixel dimensions may vary between generations. Use a custom size when you need exact pixels.
• auto lets the model choose the size (about 1.5 megapixels). It works with any resolution and is billed at the 1K tier.
• Custom WIDTHxHEIGHT (use x or X, for example 2304x1536) works with any resolution. Both edges must be multiples of 16, the longest edge must not exceed 3,840 pixels, the aspect ratio must not exceed 3:1, and total pixels must be between 655,360 and 8,294,400.
• A custom size is billed at the tier that matches its total pixels, or at the selected resolution if that is higher:

​

Examples

• Edit with a reference image: replace the background at 2K without size, so the output keeps the reference image’s aspect ratio.
• Mask edit: change only the masked area, with size: auto.
• Transparent background: remove the background, WebP, 4K, xhigh quality.
• Custom size: place the product in a panoramic scene at 2304x1536, 2K, medium quality.
• Compressed JPEG: 3:2, medium quality, output_format: jpeg, output_compression: 80.

​

Notes

• Each request returns exactly one image; n other than 1 returns 400. Fields not listed above return 400. model and callback_url belong at the request root, not inside input.
• A transparent background with output_format: jpeg returns 400. output_compression with output_format: png also returns 400.
• An empty image_urls array, more than 16 images, or a URL that is not a valid HTTP(S) URL returns 400.
• Credits depend on quality, the billing tier (resolution, or the custom-size tier described in Size) and the number of reference images. complexity and moderation do not change the price. The per-image maximum is reserved at submission; when the task finishes you are charged for actual token usage, up to that maximum, and the rest is refunded. A failed task is refunded in full. See the pricing page for current rates.