GPT Image 2.5 Flare:official is designed for image creation tasks that require quickly exploring visual directions. It can generate images from text and modify existing assets using reference images. It is suitable for ad drafts, product scenes, interface concepts, and content illustrations, with a focus on moving creation from a brief to comparable visual options. Compared with the high-fidelity, fine-control focus of the Sunburst model in the same series, Flare places greater emphasis on everyday generation and iteration efficiency.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
Flare is positioned for fast creation. The quantities, dimensions, file requirements, and response formats above are the request specifications for the corresponding platform interface.
Write the subject, environment, color palette, lighting, and layout intent into the prompt to explore advertising visuals, illustrations, or product scenes. Compare different compositions first, then refine the selected direction; clearly specifying negative space and the intended use of the image makes it easier to organize the creative process than providing only a style name.
When you already have product photos or design sketches, you can submit reference images and describe the background, colors, or scenes that need adjustment. Editing instructions should clearly state both “what to change” and “what to preserve”—for example, preserve the subject, camera angle, and lighting while replacing only the environment—so generated results can be integrated into an existing design workflow.
When a specific area needs focused changes, upload the original image together with an Alpha mask, allowing transparent areas to define the editable range. After completion, choose PNG, JPEG, or WebP according to the intended use, and obtain a link or Base64 data for easy integration with asset storage, manual review, and subsequent layout steps.
Enter the campaign theme, product selling points, visual tone, and placement dimensions to generate promotional visuals for review. Compare subject placement, backgrounds, and negative space first, then select a version suitable for adding a headline; final copy and brand elements can be refined in design tools and delivered as an advertising background or social content image.
Use product photos as references to try seasonal backgrounds, studio environments, or different color combinations. Clearly specify in the prompt that the product shape and viewpoint should be retained, and limit the parts that need to change. The deliverables are scene candidate images for choosing a direction; packaging text and structural details should be compared with the original image item by item.
Provide the page purpose, layout sketches, color preferences, or scene descriptions to create landing-page visual concepts, application interface mockups, or storyboard frames. The results are suitable for communicating mood, hierarchy, and composition, but do not constitute a functional interface or video; after selecting a direction, proceed to interaction design or motion production.
When the task focuses on quickly seeing multiple visual directions, such as campaign concepts, content illustrations, or product background drafts, Flare's speed-oriented approach better fits the workflow pace. If the evaluation focus shifts to preserving reference image details, fine editing, and high-fidelity control, compare GPT Image 2.5 Sunburst. It is recommended to evaluate results using the same assets and revision goals rather than judging only by model names.
gpt-image-2.5-flare:official and Flare without the suffix share the same core creative positioning; :official is not a new standalone image version. When using this endpoint, explicitly provide the full ID: use the generation endpoint to create from text, and the editing endpoint to modify existing assets; when local control is needed, upload the original image and mask together.
Provide prompt for generation; provide both image and revision instructions for editing. Clearly specify the original text, subject preservation requirements, and target aspect ratio.
Call the image generation or image editing endpoint and explicitly specify model=gpt-image-2.5-flare:official; use auto or WIDTHxHEIGHT for size, and generate one image first before evaluating. Set masks, quality, and file format according to this endpoint guide.
Read the URL or Base64 image according to the response format, save task_id asynchronously before querying results; check text, reference details, and alpha channels, and record usage according to the current Pricing rules.
Create a welcome-page concept image for a health tracking app, with a simple card layout, the title “Make Today Count,” a green-and-white primary color scheme, and a walking person as the main illustration.
First check the text and spatial hierarchy, then decide whether refinement is needed; record usage and the result, as an image is not equivalent to directly runnable page code.
Flare is better suited for everyday generation, draft comparisons, and rapid iteration; Sunburst focuses more on high fidelity and fine control. If product appearance or reference-material details are the main acceptance criteria, compare Sunburst; if you first need to determine composition and creative direction, start with Flare.
Explicitly set model to gpt-image-2.5-flare:official. For text-to-image generation, use /openai/images/generations and submit model and prompt; to edit an existing image, use /openai/images/edits, add image and editing instructions, and avoid relying on the default model.
Upload the original image and mask together via multipart. The mask must be a PNG with an Alpha channel, the same dimensions as the first original image, and no larger than 4MB; transparent areas indicate where editing is allowed. The prompt should still describe the complete target image and clearly specify what must remain unchanged.
The platform supports requesting 1–10 images with n; if b64_json is selected, only 1 image is supported. Read the URL or Base64 data from data in synchronous results; for asynchronous processing, set callback_url, save task_id first, then receive the completed result.
gpt-image-2.5-flare:official is billed based on actual Token usage, not a fixed per-image fee. Generation and editing have different input compositions, and reference images are also counted as editing input; pre-request estimates may differ from final usage, so actual usage and usage records prevail.
Choose a development approach or application suited to this model and follow its setup guide.
Python SDK Integration Guide
SDK installation, authentication and examplesTypeScript SDK Integration Guide
SDK installation, authentication and examples
