gpt-image-2 is OpenAI's image creation model, suitable for turning text requirements into posters, infographics, product visuals, and character designs, and can also adjust colors, backgrounds, and image content based on existing images. On this platform, it provides separate generation and editing entry points, making it suitable both for exploring visual directions from scratch and for continuously revising existing assets, delivering images rather than ordinary Q&A text.
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.
The values above are the call ranges for this platform's image entry point. Native creation capabilities and specific parameter support for each public invocation ID should be understood separately.
gpt-image-2 is suited to placing titles, slogans, labels, and visual subjects within the same image, organizing layouts around posters and infographics. Prompts can separately specify text content, hierarchy, placement, and whitespace, making results easier to control than simply describing an art style; spelling, numbers, and text arrangement should still be checked before final delivery.
When editing, you can submit an image and revision instructions directly, such as preserving the cup's shape, camera angle, and shadows while changing only the cup color and background. Describe separately what must be retained and what may change; this is suitable for product recoloring, scene replacement, and visual concept iteration, but generative editing should not be understood as pixel-perfect locking.
Multi-image input is suitable for combining subject photos, style samples, and composition references to create product concepts or character designs. It is recommended to specify what each image is responsible for: which determines appearance, which determines color palette, and which is for layout reference only, preventing visual requirements from different materials from constraining one another and making revision goals clearer.
Enter the event theme, title, required copy, and canvas requirements to generate visual drafts containing the main subject, text hierarchy, and decorative elements. This is suitable for comparing different compositions first and then refining a chosen direction. Data in infographics should be provided and verified by you; the model handles visual expression and does not replace data proofreading or final layout review.
Upload the original product image and describe the background, color, or scene to be replaced, while clearly preserving the packaging outline, viewing angle, and key identifying features. Deliverables can be used for e-commerce key visuals, advertising concepts, and packaging presentation drafts; when labels, trademarks, and product details are involved, compare them item by item with the original materials before publishing.
Provide character references and style requirements to generate expressions, clothing, or multi-view design images, and to explore a series of visuals around the same theme. Prioritize clearly defining sections and viewpoints in one image, then refine over successive rounds; when creating across images, keep carrying the reference images and check whether facial features, accessories, and clothing details remain consistent.
If a task involves text generation, reference-image revisions, and layout exploration at the same time, gpt-image-2 is suitable as a starting point. The related gpt-image-2.5-flare focuses on generation speed, while gpt-image-2.5-sunburst focuses on high fidelity and fine control; choose based on delivery goals rather than assuming that every task must use a different model solely based on the version number.
Use /openai/images/generations for creation from scratch, and /openai/images/edits for revisions based on an original image, explicitly specifying model. When a mask is needed to limit the area, you can choose the file-upload method for gpt-image-2:official; :official and :reverse are public calling ID variants and do not represent different native generations.
Provide prompt when generating; provide both image and revision instructions when 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; use auto or WIDTHxHEIGHT for size, and generate one image 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 the result; check text, reference details, and the alpha channel, and record usage according to the current Pricing rules.
Create a concise infographic based on the provided product image: center the product, place two provided short feature statements on each side, use a blue-and-white color scheme, maintain clear hierarchy, and do not add parameters that were not provided.
First verify the short feature statements against the actual product information, then review the text, arrows, and relationships between components; the generated result is an image, not an editable vector file.
It can generate titles, slogans, labels, and the main visual subject together. Provide accurate copy, and specify the title position, font-size hierarchy, and whitespace requirements. It is suitable for establishing a complete visual direction, but cannot guarantee that all text will be accurate in one attempt; spelling, numbers, and layout should still be checked before publishing.
Submit the original image through the editing interface. In the prompt, first list the subject shape, camera angle, lighting, and background that need to be preserved, then clearly state the color to change. Avoid adding unrelated style requirements at the same time. After generation, compare the result with the original image to check details rather than assuming that unmentioned areas will never change.
The image field in a JSON editing request can be a single URL or an array of up to 16 URLs; local images should be uploaded using multipart. It is best to explain the purpose of each image in the prompt, for example, which image to reference for subject appearance, color palette, and composition, to reduce ambiguity between assets.
Use size=auto to let the model choose dimensions based on the image requirements, or specify WIDTHxHEIGHT to control pixels. n can be set to 1–10, which is suitable for comparing visual options; if using the b64_json response format, only 1 image is supported. When specifying dimensions, width, height, and total pixel limits must also be met.
gpt-image-2 here is for directly generating or editing images and should not be regarded as the full ChatGPT product functionality. A single image request will not automatically perform web searches or multi-step asset orchestration; for long tasks, you can set callback_url, save the task_id first, and then receive the completed image result.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗↗Pricing
$0.0105· per generation
You pay$13,584for 142,800 credits · $0.0951 / credits46% OFF
You're billed in credits. Pick a top-up tier to see its per-credit rate — the larger the top-up, the cheaper each credit.
This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs