OpenAI · Image Creation & Source Revisions
Create an image from a brief or revise one selected source with GPT Image 2.5 Sunburst. Review the generated candidate before replacing your original.
Compare models for product images →These are conservative provider estimates and Promptha credits at the stated example settings. Your actual request receives a fresh quote.
Use a precise description of the subject, required wording and details to preserve. Compare the whole candidate with the source after editing.
The local integration supports one output, one edit source, PNG, three quality settings and standard aspect ratios. Availability depends on this environment; broader provider controls are not enabled.
Check lettering, layout and unchanged details. A successful request does not guarantee accurate text or faithful preservation.
A fictional unmarked coffee bag on a pale stone counter, a ceramic cup behind it, soft side light and believable paper texture. Leave open space to the right. No readable labels.
Illustrative prompt. Select the approved facts when they are needed.
Read the step-by-step guide →Shift this selected daylight scene toward warm early evening. Keep the table as the main subject and retain the overall framing where possible. Avoid adding readable signs.
Illustrative prompt. Select the exact eligible source image before submitting.
Read the step-by-step guide →Follow the catalog relationships for this model. Compatibility is separate from a tested output.
Open connected model detail →Provider documentation and input schema checked on . Task suggestions are editorial guidance, not a measured comparison. Price estimates are available for registered actions. Workspace activation and live output quality remain separate verification; use the fresh request quote for the actual cost.
Read the GPT Image 2.5 Sunburst provider documentation →Already working in Promptha? Open your workspace to see the models currently enabled for your request.
GPT Image 2.5 Sunburst uses the gpt-image-2-5-sunburst action for image creation. This action starts from the written request without a required image source. The configured request allows up to 1 output. Supported aspect choices: 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16.
Separate the main subject, setting, light and composition. For an edit, bind the exact eligible image; supporting inspiration is a different role. Choose a supported aspect ratio around the eventual slot, leaving space for native text when the design needs it.
Follow the product images workflowGPT Image 2.5 Sunburst uses the gpt-image-2-5-sunburst-edit action for image editing. This action requires one eligible source image and accepts at most 1. The configured request allows up to 1 output. Supported aspect choices: 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16.
Separate the main subject, setting, light and composition. For an edit, bind the exact eligible image; supporting inspiration is a different role. Choose a supported aspect ratio around the eventual slot, leaving space for native text when the design needs it.
Follow the image editing workflowOpen the complete image and inspect subject placement, object proportions, material detail and lettering. Compare product-specific work with its approved source. A credible scene can still change an important detail; choose an exact version before using it in a native layout.
Does the frame serve the actual product or layout job? Keep the source and brief fixed when comparing models. Public model descriptions explain intended uses and input contracts; they are not live benchmarks or evidence of a universal quality ranking.
Compare exact saved versionsThe image editing route requires an eligible selected source. Inspect the exact image and the exposed controls before preparation; other actions have their own input contracts.
Registered adapters and pricing policy do not guarantee activation. Check enabled models, permissions and current lifecycle in your workspace. An unavailable model needs an explicit compatible replacement and a fresh quote.