GPT Image 2.5 Sunburst: Precise Campaign Image Editing

openai/gpt-image-2.5/sunburst/edit

GPT Image 2.5 Sunburst focuses on precise image revisions for campaign artwork and polished product photos. Use reference images and detailed instructions to refine a chosen direction, with output up to 4K.

Idle

Price per output image by quality x resolution (1K / 2K / 4K): low $0.019 / $0.029 / $0.039, medium $0.033 / $0.049 / $0.079, high $0.099 / $0.159 / $0.279, xhigh $0.169 / $0.279 / $0.489, max $0.369 / $0.609 / $1.009. Each additional reference image beyond the first adds $0.012.

Introduction to GPT Image 2.5 Sunburst

GPT Image 2.5 Sunburst is for the stage when small visual decisions matter: the treatment of a product surface, the lighting around a hero image, or a revision to an approved campaign composition. This RunComfy editing page lets you describe those changes against reference images. Choose it when closer control is worth a longer wait.

GPT Image 2.5 Sunburst: Precise Editing for Final Visuals#

GPT Image 2.5 Sunburst suits image revisions with a specific creative brief. Start from your product photo or campaign artwork, explain the intended change, and use the result to assess whether the details meet your requirements.

Why choose Sunburst?#

OpenAI positions Sunburst for workflows where editing precision matters most. It is a useful candidate when you have already chosen a direction and want to work through the remaining visual decisions. Model details.

Where it fits#

Campaign finishing: Refine the scene around an approved subject before the asset goes into layout.
Product hero images: Explore a more deliberate treatment of the background, lighting, and materials around a product.
Art-directed composites: Give separate references explicit roles, such as the product, environment, and color treatment.
Detailed revisions: Work through feedback with a focused instruction for each pass and compare against the approved source.

Treat each output as an asset to review. Brand-critical lettering, exact packaging, and small details may still need manual finishing.

How to use it on RunComfy#

Upload the image you want to refine. If you use several references, name the role of each in the prompt.
Describe the final visual, including the change and the details that must remain recognizable.
Choose the aspect ratio for the delivery format. The current defaults are 1:1, 1K, medium quality, and PNG.
Generate a version and inspect it against the source. For product work, review the silhouette, label, cap, and reflections separately.
Upload the selected output for a further revision, or return to the original if an important detail has drifted.
Evaluate higher quality and resolution settings for the selected direction. Recheck the resulting image before placing final copy or publishing.

Example: Finish a product hero image#

> Use image 1 as the product reference. Keep the bottle silhouette, cap, label wording, and front-facing angle. Replace the surrounding scene with an off-white studio surface and a warm gray backdrop. Add soft light from the upper left and a restrained contact shadow. Leave the right third clear for copy. Do not add decorations or new lettering.

For a follow-up revision, identify the remaining issue specifically, such as a shadow that is too strong. Keep the source nearby for comparison.

Sunburst or Flare?#

Choose Sunburst when precise revisions are the priority. Choose Flare when rapid exploration is more useful, especially for frequent creator and social content. Sunburst takes longer according to OpenAI; that tradeoff does not guarantee that every image will be better. Release announcement.

For an existing GPT Image 2 or other-model workflow, test the same source and brief before switching. Compare how many required details survive, how much manual correction remains, and the total cost of reaching an acceptable result.

Available controls#

Control	Current options	Default
images	1 to 16 reference images	Required input
prompt	Written editing instruction	Required input
aspect_ratio	1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9, 2:1, 1:2, 3:1, 1:3, 9:21	1:1
resolution	1k, 2k, 4k	1k
quality	low, medium, high, xhigh, max	medium
output_format	png, jpeg, webp	png

Pricing#

Check the current RunComfy price for the selected settings before running. Quality, resolution, reference count, and output count can affect the total.

Prompting reference#

See the OpenAI image prompting guide for examples of reference-based editing. The controls above describe this RunComfy page; other interfaces may expose additional options.

Related Models

seedream-5.0-pro/image-to-image

Reference-guided image editing with layout, layer separation, and multilingual text

seedream-4-0/edit-sequential

Create cohesive visual sequences with precise style and continuity control.

flux-1-kontext/max/text-to-image

Advanced model with fast text control, precision edits, and consistent visual fidelity.

seedream-4-5/edit

Transform visuals with Seedream 4.5 for coherent, photoreal image creation and precise brand consistency.

ideogram-v3/text-to-image

Sync image edits, remixes, reframe, and background swaps for film.

z-image/turbo/image-to-image

High-speed image transformation with precision lighting and bilingual prompt support.

Frequently Asked Questions

What is GPT Image 2.5 Sunburst best for?

Sunburst focuses on precise image editing. It is a useful option for campaign artwork, polished product imagery, and revisions where the details matter more than rapid exploration.

How is Sunburst different from Flare?

Sunburst prioritizes tighter editing control with longer generation times. Flare is the speed-focused option for frequent iteration. Both can be useful in a project: Flare for exploring a direction and Sunburst for evaluating detailed revisions.

Can Sunburst guarantee an unchanged logo or product label?

No. Ask it to retain the relevant details, then inspect lettering, product geometry, and other brand elements against the source. Exact artwork may require a final manual correction.

How should I write a product editing prompt?

Identify the product reference, describe the surrounding scene and lighting, and list important details to retain. Specify where copy space should go. Avoid asking the model to invent missing packaging details.

How do I continue editing a previous result?

Upload that result for the next run and describe the next change. Repeat important requirements. This page submits images and a prompt for each run; it does not expose automatic conversation history.

What inputs and output settings does this page support?

Supply a prompt and 1 to 16 reference images. Choose 1K, 2K, or 4K; low through max quality; and PNG, JPEG, or WebP. The current defaults are 1:1, 1K, medium, and PNG.

Can I request a transparent background here?

This page does not expose a dedicated background or transparency setting. Selecting PNG alone does not guarantee transparency. Do not rely on this page for an alpha-channel delivery requirement without checking the output.

Is Sunburst always more expensive or better than Flare?

Do not assume either from the name alone. Check the current RunComfy price for your settings and compare both models on your actual images. Choose the one that gives you an acceptable result with less revision work.

RunComfy

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

GPT Image 2.5 Sunburst: Precise Campaign Image Editing

GPT Image 2.5 Sunburst focuses on precise image revisions for campaign artwork and polished product photos. Use reference images and detailed instructions to refine a chosen direction, with output up to 4K.

Introduction to GPT Image 2.5 Sunburst