How to Use GPT Image 2.5 Sunburst

Learn how to use GPT Image 2.5 Sunburst in CawCut for precise image creation, reference-guided edits, and polished creative work.

Updated Nov 18, 2026 · 7 min read

When an image needs careful detail and edits that preserve the rest of the composition, GPT Image 2.5 Sunburst is the precision-focused choice in CawCut's [Image Generation] node. It is built for premium creative work, with particular strength in detailed edits and polished product or campaign imagery. This guide shows how to use it for both new images and reference-guided changes.

1. GPT Image 2.5 Sunburst Strengths

StrengthApplication
Precision for detailed creative work: Sunburst offers an extra level of precision for complex visual direction and targeted edits.Create a finished campaign image, detailed product visual, or polished creative asset where small compositional choices matter.
Focused edits that preserve surrounding details: GPT Image 2.5 is designed to change the requested element while keeping the subject, layout, and other parts of the image steady.Replace a product, adjust a background, or revise a piece of copy without redesigning the entire asset.
Strong reference fidelity: The model is better at carrying recognizable subjects, distinctive features, lighting, and textures from a reference into a new composition.Place a familiar character, product, or person into a different setting or visual style while retaining the features that make the subject recognizable.
Reliable refinement across multiple edits: The model is designed to follow successive instructions while keeping earlier changes and overall image quality consistent.Develop a visual in stages—such as changing the setting, then adjusting wardrobe or lighting—without restarting from the original image each time.
Natural lighting, rich textures, and complex instruction handling: GPT Image 2.5 improves texture, lighting, and its ability to interpret detailed visual briefs.Produce material-rich product images, layered scenes, or structured creative layouts with a clearly defined visual direction.

Sunburst's precision comes with longer generation times than Flare. If you are deciding between them, see GPT Image 2.5 Sunburst vs. Flare.

2. Choose a Generation Mode

CawCut's Image Generation node supports both modes with Sunburst:

ModeChoose it whenWhat to provide
[Text to Image]You want to create a new image from a written brief.Describe the subject, composition, visual style, lighting, and intended use.
[Image to Image]You want to edit, restyle, combine, or build on existing visual material.Add the source and reference images, then specify what to change and what to preserve.

The [Image Generation] node also lets you compare multiple outputs from the same prompt and settings. For details on the mode controls and model-specific options, see Image Generation.

3. Prepare Your Prompt and References

Sunburst is most useful when you identify the important details precisely. For a new image, organize the brief around:

  • Subject: Name the main object, person, or scene and its defining features.
  • Composition: State where the subject sits in the frame, what is in the foreground or background, and what must remain visible.
  • Style and lighting: Describe materials, texture, color palette, light direction, and mood.
  • Text or layout: Quote exact wording and describe its position, hierarchy, and relation to other visual elements.
  • Output intent: Choose the aspect ratio and resolution for the image's destination, such as a product page, campaign banner, or social post.

For an edit, make the boundaries clear:

  1. Identify the source image and the specific element or region to change.
  2. Describe the edit as a visible result, such as “replace the plain wall with a warm tiled backsplash.”
  3. List important details to keep unchanged, including the camera angle, product shape, subject pose, logo, or overall palette.
  4. Avoid combining unrelated edits in one instruction. Refine one part at a time when a close match matters.

Use a clean, high-quality reference with the subject and important details visible. If your references should pass between nodes, see Image and Image Group Nodes.

4. CawCut Workflow and Settings

  1. Add an [Image Generation] node to a workflow.
  2. Choose [Text to Image] for a new image or [Image to Image] to work from an existing image.
  3. Select GPT Image 2.5 Sunburst from the model menu.
  4. Enter the prompt and connect or add source images if you are editing.
  5. Set the aspect ratio for the destination. CawCut lists landscape, portrait, square, and other wide formats, including [16:9], [9:16], [1:1], [21:9], [4:5], and [Auto].
  6. Choose [1K], [2K], or [4K] resolution. Use [Low], [Medium], [High], [XHigh], or [Max] quality according to the level of detail you need.
  7. Choose a [Filled] background for a complete scene or [Transparent] when you need a cutout-style asset, then choose how many outputs to generate.
  8. Review the credit estimate, run the node, and compare the results before selecting the version that best fits your brief.

For help passing images between connected nodes, see Understanding Nodes. If you are comparing quality options or image versions, the Image Generation guide explains those controls.

5. Limitations and Refinement

Sunburst is the slower GPT Image 2.5 variant, so it is best reserved for work where precision and polish justify the longer wait. It can interpret detailed instructions more reliably, but exact text, fine geometry, and every reference detail still need a careful review. A generated image is a visual draft until you confirm the elements that matter for your use.

If you noticeTry this
The edit changes more than requestedName the single element to change and explicitly preserve the subject, composition, lighting, and other untouched areas.
A reference subject loses defining featuresUse a clearer reference and describe the distinctive features that must carry through. Remove references that suggest a conflicting appearance.
Text or a label is inaccurateQuote the exact wording, simplify the layout, and inspect the result at full size. Use a design editor for final typography that must be exact.
The composition feels crowdedRemove secondary elements and state the focal point and where it should sit in the frame.
Iteration is taking too longTry GPT Image 2.5 Flare for faster exploration, then return to Sunburst for the precision-oriented final pass.

For CawCut model alternatives and account credit guidance, see Supported Models and Credits.

Related articles

Frequently Asked Questions

Learn how to use GPT Image 2.5 Sunburst in CawCut for precise image creation, reference-guided edits, and polished creative work.

Sunburst is designed for detailed creative work that benefits from more precision and tighter control across edits, such as polished product imagery and production-ready campaign creative.

Sunburst focuses on precision for detailed creative work and takes longer to generate. Flare is the faster option for general creation, editing, and rapid iteration. See GPT Image 2.5 Sunburst vs. Flare for help choosing.

Yes. Choose Image to Image, add the source image, and describe the change. Clearly state which parts to change and which details to preserve.

CawCut lists 1K, 2K, and 4K resolutions and Low, Medium, High, XHigh, and Max quality for Sunburst. Choose based on whether you are exploring a concept or preparing a detailed final image.

Name the specific object or area to edit, describe the desired replacement or adjustment, and explicitly say what should remain unchanged. Make one focused edit per generation when practical.

Sunburst is the precision-oriented variant and generally takes longer to generate. Choose it when detailed control matters more than turnaround time.