GPT Image 2.6 Sunburst Image Edit API

openai/gpt-image-2.6-sunburst/edit

image_urls · mask_url

GPT Image 2.6 Sunburst Image Edit turns 1–16 reference images and text instructions into a refined 1K–4K image, with masks, multi-image composites and quality up to max. Built for stronger consistency across edits, it keeps identity, materials and lighting steady while applying the requested change.

Get API Key

Input

PromptRequired649/32000

Image UrlsRequired

Drag and drop files, paste from clipboard, or provide a URL. Provide 1–16 public HTTP(S) image URLs without embedded credentials, in the order the prompt refers to them. Each image adds 2.50 credits to the upfront cap.

2/16 files added

SizeOptional

ResolutionOptional

Output FormatOptional

QualityOptional

BackgroundOptional

ModerationOptional

ComplexityOptional

Mask UrlOptional

Drag and drop files, paste from clipboard, or provide a URL. Optional single public HTTP(S) mask image. Its fully transparent area marks the region of image 1 to change. Masks are free.

OutputReady

Maximum upfront charge: 14.50 credits · $0.07250

Billed by actual tokens. Any difference is refunded after completion.

Continue using

GPT Image 2.6 Sunburst Text to Image GPT Image 2.6 Flare Image Edit GPT Image 2.5 Sunburst Image Edit

Examples

Edit image 1 using the supplied mask. Inside the masked area, give the violin body its first coat of varnish: the spruce top and maple ribs become a translucent warm golden-amber with a soft wet sheen, while the flame figure in the maple and the fine spruce grain stay clearly visible through the varnish. Keep the luthier's hands, fingers and the small brass thumb plane exactly as they are, in the same positions. Keep the grey wool cloth, the workbench, shavings, the separate carved neck and scroll, the luthier, the hanging violins and the window light unchanged. No text, no logos, no watermark.

Use image 1, the screen-printed travel poster, as the only reference. Turn it into an autumn edition. Change the line under the headline to read exactly "AUTUMN FERRIES" and the bottom band to read exactly "DAILY CROSSINGS · SEPTEMBER – NOVEMBER". Keep the headline "PORT ALDER" exactly as it is. Shift the palette to autumn: deep amber and rust sky, golden and russet trees on the shore, slightly darker teal water. Keep the same typefaces, letter sizes, layout, ferry, lighthouse, gulls, hills, paper grain and print texture. These three lines remain the only text. No logos, no watermark.

REST API

Quick Start

This flow submits one valid GPT Image 2.6 Sunburst Image Edit request, captures its task_id, and reads image URLs after the asynchronous task ends.

Connect to the Vidgo API

Create an API key, keep it on the server, and send it as Authorization: Bearer VIDGO_API_KEY.

Endpoint: POSThttps://api.vidgo.ai/api/generate/submit
Authentication: Authorization: Bearer VIDGO_API_KEY

Submit one image task

Send the current model ID, prompt, and ordered image URLs to the unified submit endpoint. The example keeps only first-call fields.

Submit request

curl --request POST \
  --url "https://api.vidgo.ai/api/generate/submit" \
  --header "Authorization: Bearer $VIDGO_API_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "model": "openai/gpt-image-2.6-sunburst/edit",
  "input": {
    "prompt": "Replace the sky with a starry night sky. Keep everything else unchanged.",
    "image_urls": [
      "https://example.com/image.png"
    ],
    "resolution": "1K",
    "output_format": "png",
    "quality": "medium",
    "complexity": "low"
  }
}'

Wait for the image result

Read data.task_id and poll only while status is not_started or running. Stop on both finished and failed.

Track status

GET https://api.vidgo.ai/api/generate/status/{task_id}

Poll every 2–5 seconds and back off on 429/5xx. Or provide top-level callback_url to receive a terminal-state callback.

not_startedrunningfinishedfailed

Submit response

{
  "code": 200,
  "data": {
    "task_id": "task-example",
    "status": "not_started",
    "created_time": "2026-09-25T08:00:00"
  }
}

Finished response

{
  "code": 200,
  "data": {
    "task_id": "task-example",
    "status": "finished",
    "created_time": "2026-09-25T08:00:00",
    "files": [
      {
        "file_url": "https://example.com/result.png",
        "file_type": "image"
      }
    ]
  }
}

Complete runnable polling example

Expand for the same payload end to end, including business-code, missing-task-ID, failed-state, unexpected-state, and network-error handling.

Complete runnable polling example

set -euo pipefail
: "${VIDGO_API_KEY:?Set VIDGO_API_KEY in your environment}"

REQUEST_BODY=$(cat <<'JSON'
{
  "model": "openai/gpt-image-2.6-sunburst/edit",
  "input": {
    "prompt": "Replace the sky with a starry night sky. Keep everything else unchanged.",
    "image_urls": [
      "https://example.com/image.png"
    ],
    "resolution": "1K",
    "output_format": "png",
    "quality": "medium",
    "complexity": "low"
  }
}
JSON
)

SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  --request POST \
  --url "https://api.vidgo.ai/api/generate/submit" \
  --header "Authorization: Bearer $VIDGO_API_KEY" \
  --header "Content-Type: application/json" \
  --data "$REQUEST_BODY")

TASK_ID=$(printf '%s' "$SUBMIT_RESPONSE" | jq -r '.data.task_id // .task_id // empty')
if [ -z "$TASK_ID" ]; then
  printf 'Submit response did not include task_id:
%s
' "$SUBMIT_RESPONSE" >&2
  exit 1
fi

while true; do
  STATUS_RESPONSE=$(curl --silent --show-error --fail-with-body \
    --url "https://api.vidgo.ai/api/generate/status/$TASK_ID" \
    --header "Authorization: Bearer $VIDGO_API_KEY")
  STATUS=$(printf '%s' "$STATUS_RESPONSE" | jq -r '.data.status // .status // empty')

  case "$STATUS" in
    finished)
      printf '%s' "$STATUS_RESPONSE" | jq -r '(.data.files // .files // [])[]?.file_url'
      break
      ;;
    failed)
      printf '%s' "$STATUS_RESPONSE" | jq -r '.data.error_message // .error_message // "Generation failed"' >&2
      exit 1
      ;;
    not_started|running)
      sleep 2
      ;;
    *)
      printf 'Unexpected task status: %s
' "$STATUS" >&2
      exit 1
      ;;
  esac
done

Request Parameters

This table lists only public fields inside input. See Quick Start for top-level model and optional callback_url usage.

Field	Type	Required	Default	Description
input.prompt	string	Yes	–	Edit instructions, non-blank, up to 32,000 characters. Refer to references by order, such as image 1.
input.image_urls	string[] \| string	Yes	–	Provide 1–16 public HTTP(S) image URLs without embedded credentials, in the order the prompt refers to them. Each image adds 2.50 credits to the upfront cap.
input.size	string	No	Follows image 1	When omitted, the output follows the first reference image's aspect ratio. Ratios: auto, 1:1, 2:3, 3:2, 4:3, 3:4, 4:5, 5:4, 16:9, 9:16, 21:9, or custom WIDTHxHEIGHT. auto works at any resolution and is billed at 1K. Custom WIDTHxHEIGHT works at any resolution: positive multiples of 16, maximum edge 3840, ratio at most 3:1 and 655360–8294400 pixels, billed at the higher of the selected resolution and its pixel tier (up to 1,572,864 px 1K, up to 4,194,304 px 2K, above that 4K).
input.resolution	string	No	1K	1K, 2K or 4K. Default 1K. 1K, 2K, 4K.
input.output_format	string	No	png	png, jpeg or webp. Default png. png, jpeg, webp.
input.quality	string	No	high	low, medium, high, xhigh or max. Default high. Quality and the billing tier set the upfront cap; final billing uses actual tokens. low, medium, high, xhigh, max.
input.background	string	No	auto	auto, transparent or opaque. Default auto. transparent requires png or webp. auto, transparent, opaque.
input.output_compression	integer	No	–	Integer from 0 to 100 for jpeg or webp output. Leave empty for the default compression.
input.moderation	string	No	auto	auto or low. Default auto; low applies less restrictive filtering. Does not change the price. auto, low.
input.complexity	string	No	high	low, medium or high. Default high. Sets how much the model plans before rendering and does not change the price; low suits most edits. low, medium, high.
input.mask_url	string	No	–	Optional single public HTTP(S) mask image. Its fully transparent area marks the region of image 1 to change. Masks are free.

Response Fields

The submit response identifies the task. The status response supplies image files or failure details at a terminal state.

Field	Type	Description
code	integer	Business response code; 0 or 200 indicates success.
message	string	Optional human-readable response message.
data.task_id	string	Unique task ID used for status queries after submission.
data.status	string	not_started, running, finished, or failed.
data.created_time	string	Task creation timestamp.
data.progress	integer	Progress from 0–100 in status responses.
data.files[]	array	Generated files in returned order when finished.
data.files[].file_url	string	Direct generated-image URL.
data.files[].file_type	string	Image results are identified as image.
data.files[].watermark_url	string \| null	Optional watermarked-file URL when present.
data.error_message	string \| null	Readable failure reason for failed tasks.

Task Lifecycle

Only the first two states may be polled again. finished and failed are terminal.

not_started
Accepted and queued; wait before checking again.
running
Image generation is in progress; continue polling.
finished
Successful terminal state; validate files and read image URLs.
failed
Failure terminal state; read error_message and stop.

Polling and Errors

AuthenticationKeep a valid VIDGO_API_KEY on the server and send requests with Bearer authentication.
Validation400 indicates an invalid model ID, field, enum, length, or image count.
Polling intervalStart at 2–5 seconds, use exponential backoff for 429/5xx, and enforce an overall timeout.
Terminal statesRead image URLs from files after finished, or error_message after failed. Both states end polling.
Callback optionProvide a public top-level callback_url to receive a flat task object after finished or failed.
Network errorsDistinguish request timeout, network failure, and task failure. Retry the same query only when safe.

Model Specifications

Specification	Value	Details
Model	openai/gpt-image-2.6-sunburst/edit	Edit reference images using text instructions.
size	auto / 1:1 / 2:3 / 3:2 / 4:3 / 3:4 / 4:5 / 5:4 / 16:9 / 9:16 / 21:9 / WIDTHxHEIGHT	When omitted, follows the first reference image's aspect ratio.
Output	1 image per request · png / jpeg / webp	jpeg and webp accept output_compression from 0 to 100.
complexity	low / medium / high	Planning effort before rendering. Default high; does not change the price.

GPT Image 2.6 Sunburst Image Edit API documentation

GPT Image 2.6 Sunburst Image Edit

GPT Image 2.6 Sunburst Image Edit uses the precision editing model in OpenAI’s Images 2.6 release from November 2026. OpenAI states that at medium or high quality Sunburst delivers higher-quality edits and stronger consistency across edit rounds. It accepts up to 16 ordered references and an optional mask for retouching, product placement and text changes on existing artwork, and the output follows the first image’s aspect ratio unless you set size.

Why Choose GPT Image 2.6 Sunburst Image Edit?

Consistency across edit roundsOpenAI positions Sunburst for stronger consistency across edits at medium and high quality, keeping faces, materials and lighting steady over several passes.
Composites from multiple referencesPlace a product or character from one image into the scene of another, matched to its light and scale.
Text changes on existing designsRewrite headlines or dates on posters and packaging while typefaces and layout carry over, as in the Port Alder autumn edition.
Masked finishing workConfine a change such as varnish, fabric or color to a masked region of the source image.
Aspect ratio follows image 1Leave size empty to keep the source frame, and set it only when you need a new one.
Planning and price controlcomplexity adjusts planning at the same price. The upfront cap adds 2.50 credits per reference, and the unused part is refunded.

Parameters

Parameter	Requirement	Description
prompt	Required	Edit instructions, non-blank, up to 32,000 characters. Refer to references by order, such as image 1.
image_urls	Required	Provide 1–16 public HTTP(S) image URLs without embedded credentials, in the order the prompt refers to them. Each image adds 2.50 credits to the upfront cap.
size	Optional	When omitted, the output follows the first reference image's aspect ratio. Ratios: auto, 1:1, 2:3, 3:2, 4:3, 3:4, 4:5, 5:4, 16:9, 9:16, 21:9, or custom WIDTHxHEIGHT. auto works at any resolution and is billed at 1K. Custom WIDTHxHEIGHT works at any resolution: positive multiples of 16, maximum edge 3840, ratio at most 3:1 and 655360–8294400 pixels, billed at the higher of the selected resolution and its pixel tier (up to 1,572,864 px 1K, up to 4,194,304 px 2K, above that 4K). Default`Follows image 1`
resolution	Optional	1K, 2K or 4K. Default 1K. Default`1K2K4K`
output_format	Optional	png, jpeg or webp. Default png. Default`pngjpegwebp`
quality	Optional	low, medium, high, xhigh or max. Default high. Quality and the billing tier set the upfront cap; final billing uses actual tokens. Default`highlowmediumxhighmax`
background	Optional	auto, transparent or opaque. Default auto. transparent requires png or webp. Default`autotransparentopaque`
output_compression	Optional	Integer from 0 to 100 for jpeg or webp output. Leave empty for the default compression.
moderation	Optional	auto or low. Default auto; low applies less restrictive filtering. Does not change the price. Default`autolow`
complexity	Optional	low, medium or high. Default high. Sets how much the model plans before rendering and does not change the price; low suits most edits. Default`highlowmedium`
mask_url	Optional	Optional single public HTTP(S) mask image. Its fully transparent area marks the region of image 1 to change. Masks are free.

How to Use

Order the referencesPut the image you are editing first, then any objects or style references, up to 16 in total.
Describe the target stateSay what the result should look like and list what must stay: identity, pose, typefaces, lighting.
Mask the region when neededAdd a mask whose transparent area covers the part to change.
Pick quality and sizeUse medium or high quality for consistent results; leave size empty to keep image 1’s ratio.
Run, review and continueRead the result when the task finishes and use it as image 1 for the next round.

Usage and Pricing

Billed by actual text input, image input and image output tokens. 1 credit = $0.005.

An upfront cap is held at submission from quality and billing tier, from 5.50 credits at low 1K to 142.34 credits at max 4K, plus 2.50 credits per reference image; masks are free. The cap is the most a task can cost: after completion, actual tokens are settled and the difference is refunded, and failed tasks are refunded in full.

Usage and Pricing
Usage	Credits / 1M tokens	USD / 1M tokens
Text input	800	$4.00
Image input	1,280	$6.40
Image output	4,800	$24.00

Best Use Cases

Retouching and finishingApply varnish, fabric changes or color grading inside a mask, as in the luthier example.
Product and character placementInsert an object from one image into a scene from another, like the felt fox in the reading room.
Seasonal and localized design versionsSwap dates, taglines and palettes on posters and packaging while the layout stays.
Multi-round art directionRefine a hero image over several passes with one change per round.

Pro Tips

Describe the target state rather than the operation, for example “the spruce top becomes translucent golden amber”.
Quote replacement text exactly and name the words that stay unchanged.
When combining images, say which one supplies the scene and which supplies the object, and size the object against something in the scene.
Use medium or high quality when consistency across rounds matters, and reuse each accepted output as the next image 1.

Notes

image_urls takes 1–16 public HTTP(S) URLs; mask_url takes one.
Each request returns one image. Omit size to follow image 1’s aspect ratio.
For prompt-only generation, use GPT Image 2.6 Sunburst Text to Image.
Tasks run asynchronously: submission returns a task_id, and the finished status carries the image URL.

Related Models

GPT Image 2.6 Flare Text to Image preview

GPT Image 2.6 Flare Text to ImageGPT Image 2.6 Flare Text to Image turns text prompts into 1K–4K images with fast low-quality drafts, legible poster typography and transparent backgrounds. It places headlines, subjects and props where the prompt puts them, so one prompt carries a concept from quick draft to final quality.

GPT Image 2.6 Flare Image EditGPT Image 2.6 Flare Image Edit turns 1–16 reference images and text instructions into a new 1K–4K image, with optional masks, transparent cutouts and adjustable planning effort. It keeps the first image’s aspect ratio by default, and keep-and-change instructions let it recolor or relight a subject while the rest holds.

GPT Image 2.6 Sunburst Text to Image preview

GPT Image 2.6 Sunburst Text to ImageGPT Image 2.6 Sunburst Text to Image turns detailed text prompts into 1K–4K images with fine material texture, crowded multi-element layouts and quality up to max. It keeps long scene descriptions in order, placing subjects, props and light where the prompt puts them.

Ideogram Vector 1.0 Text to Vector preview

Ideogram Vector 1.0 Text to VectorIdeogram Vector 1.0 Text to Vector turns a written prompt into SVG artwork for logos, icons and illustrations, with up to five reference images and 17 aspect ratios plus auto. Results are true vector files that scale cleanly from favicon to billboard.

Ideogram Vector 1.0 Image to Vector preview

Ideogram Vector 1.0 Image to VectorIdeogram Vector 1.0 Image to Vector turns one raster image into an SVG file, with a detailed default mode and a simple mode that uses fewer shapes. It follows the shapes and colours of the source so logos, sketches and illustrations become scalable artwork.

FLUX 3 Klein Text to ImageFLUX 3 Klein Text to Image turns prompts into product shots, scenes and illustrations at 1K, 2K or 4K, with seven presets, exact custom sizes and up to four images per call. It keeps per-image cost low so you can iterate widely before rendering finals.

FLUX 3 Klein EditFLUX 3 Klein Edit turns one to ten input images and a written instruction into edited or combined results at 1K, 2K or 4K, priced the same as generation. It keeps the subjects you name from each image while changing backgrounds, outfits, materials or style.

Nano Banana 2.1 Text to ImageNano Banana 2.1 Text to Image turns text prompts into 1K–4K images with text layout rendering and panoramic compositions. It preserves perspective and lighting while eliminating tiling artifacts in panoramic aspect ratios.

Nano Banana 2.1 EditNano Banana 2.1 Edit uses 1–10 reference images for image blending and local edits, with consistent characters and sharp redraws. It preserves character appearance and lighting as elements change, delivering crisp results at up to 4K resolution.

Ideogram 4.5 Text to ImageIdeogram 4.5 Text to Image turns written prompts into posters, wordmarks and text-rich images, with three quality tiers and 1–8 images per request. It keeps headlines and short copy legible while you pick the aspect ratio or one of 36 exact pixel sizes.

Ideogram 4.5 EditIdeogram 4.5 Edit turns one source image and a written instruction into a revised image, with optional masks, up to four reference images and high edit precision. It keeps the areas you did not ask to change, so repeated edits stay close to the original.

Qwen Image 2.1 Text to ImageQwen Image 2.1 Text to Image turns text into native 1K and 2K images with RGBA transparency and Chinese-English typography. It preserves lighting, composition, and aspect ratios while delivering clean transparent layers that need no background removal.

Qwen Image 2.1 EditQwen Image 2.1 Edit turns 1-10 reference images and instructions into precise edits, with multi-image feature fusion and masked inpainting. It preserves facial features and product materials while replacing backgrounds naturally and retaining every pixel outside the mask.

Seedream 5.0 Flash Text to Image preview

Seedream 5.0 Flash Text to ImageSeedream 5.0 Flash Text to Image turns text prompts into high-definition 1K to 2K images, with Chinese and English text layout and 8 aspect ratios. It closely follows complex prompts while providing fine control over lighting and composition.

Seedream 5.0 Flash Image EditSeedream 5.0 Flash Image Edit combines 1 to 10 reference images with a prompt to create new work at up to 2K resolution. It closely preserves faces, clothing and composition from the source images while naturally transforming their style.

GPT Image 2.5 Flare Text to Image preview

GPT Image 2.5 Flare Text to ImageGPT Image 2.5 Flare Text to Image turns text prompts into detailed 1K–4K images, with transparent backgrounds and low latency. It combines fast responses with natural lighting while avoiding structural distortion in large images.

GPT Image 2.6 Sunburst Image Edit API FAQ

What is the GPT Image 2.6 Sunburst Image Edit API?

GPT Image 2.6 Sunburst Image Edit is an OpenAI model for precise, instruction-based image editing. It takes 1–16 reference images and an optional mask and returns one 1K, 2K or 4K image, handling retouching, composites and text changes with quality up to max. As the precision model of OpenAI’s Images 2.6 release, it keeps stronger consistency across edit rounds at medium and high quality and follows the first image’s aspect ratio by default. You can call it programmatically or try it from the playground above.

Does GPT Image 2.6 Sunburst Image Edit stay consistent across rounds?

OpenAI states that at medium or high quality Sunburst keeps stronger consistency across edits. To build on that, feed each accepted result back as image 1, make one change per round and list what must stay the same.

Can GPT Image 2.6 Sunburst Image Edit change text in an image?

Yes. Quote the replacement words exactly and name the words that stay. The Port Alder example rewrote two lines and shifted the palette to autumn while the headline, typefaces and layout carried over, using complexity low at medium quality.

How do I combine two images with GPT Image 2.6 Sunburst Image Edit?

Send the scene as image 1 and the object as image 2, then say where the object goes and how large it is. The reading room example placed the felt fox from image 2 beside a lamp in image 1 and, with size left out, kept image 1’s 16:9 frame.

How do masks work in GPT Image 2.6 Sunburst Image Edit?

Send one mask image in mask_url. Its fully transparent area marks the region of image 1 to change, such as the violin body in the luthier example, and the prompt describes the new look. Masks add nothing to the price.

Which quality suits GPT Image 2.6 Sunburst Image Edit?

Medium or high for most edits, since OpenAI ties Sunburst’s edit consistency to those tiers, and xhigh or max for 4K finals. Lower complexity keeps simple edits focused at the same price.

How is GPT Image 2.6 Sunburst Image Edit billed?

By actual tokens: 800 credits per 1M text input tokens, 1,280 per 1M image input tokens and 4,800 per 1M image output tokens, with 1 credit = $0.005. The upfront cap is the quality-and-tier price plus 2.50 credits per reference, for example 12.00 credits for one image at medium 1K; unused credits are refunded.

Should I choose GPT Image 2.6 Sunburst or Flare Image Edit?

Choose Sunburst for final retouching, composites and multi-round edits where consistency matters. Choose Flare for fast recolors, cutouts and background swaps. Both accept the same fields at the same prices.