OpenAI · Image Drafts & Whole-Image Edits
Explore an image draft or change a selected image with GPT Image 2.5 Flare. Keep the exact source available for comparison.
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.
Start with a clear subject and composition, then describe one concrete revision. The saved request records the model and selected settings.
The local integration starts at low quality with one PNG output. Medium and high quality are explicit choices; multiple-source and selected-region editing are not enabled.
Inspect edges, text and subject consistency after every edit. Quality labels describe provider settings, not a measured ranking against other models.
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 Flare provider documentation →Already working in Promptha? Open your workspace to see the models currently enabled for your request.
GPT Image 2.5 Flare uses the gpt-image-2-5-flare 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 Flare uses the gpt-image-2-5-flare-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.