DALL·E 2 is OpenAI's second-generation image creation model, capable of transforming natural-language descriptions into photorealistic images or artwork, with a focus on combining subject concepts, visual attributes, and artistic styles. It is suitable for concept sketches, illustration exploration, and editing experiments with existing images. On this platform, it can be used separately through the image generation and editing entry points, delivering image results rather than ordinary conversational answers. OpenAI's current official model documentation lists DALL·E 2 as a deprecated model. For existing projects, first confirm its current availability on this platform in the console; for new projects, prioritize comparing current models such as the GPT Image series.
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.
Native creation methods describe model capabilities; this platform uses the generation or editing endpoints, while quantity and result format are API parameters.
DALL·E 2 can place different concepts, attributes, and styles into a single image, such as adding a fantasy setting to everyday objects or expressing the same theme in both realistic and painterly styles. When describing, first clarify the subject, then add the scene, materials, and style, making it easier to explore visual approaches around the same creative direction.
In addition to generating new images from text, DALL·E 2 also supports inpainting, outpainting, and variation creation. In an image-editing workflow, you can provide an original image and a description of the changes, expressing what you want to adjust. It is important to distinguish the model's creative capabilities from specific operating methods, and not treat text instructions as pixel-level locking tools.
Generation and editing use separate endpoints, making it suitable to split drafting and modification into two steps. Applications can receive image URLs or choose to process image data in Base64 form; task-based responses provide task_id. This makes it possible to build preview, selection, and subsequent saving workflows around image results, rather than packaging them as chat responses.
Enter a character, environment, and presentation style, for example, “small animals in a forest in the rain, soft watercolor,” to first obtain theme sketches, then continue exploring by adjusting color tones or scene descriptions. The deliverables are suitable for discussing illustration directions and for creative reference; when precise character specifications are needed, inspect each image individually rather than assuming repeated generations will be completely consistent.
Write product concepts, spatial atmospheres, or story imagery as image descriptions, specifying which elements are the subject and which are the background, then generate candidate images. It is suitable for comparing visual expressions before formal design and using images to help teams discuss directions; final brand text, standardized layout, and precise dimensions can be completed during the design production stage.
Provide an accessible original image URI and modification goals to the editing endpoint, such as exploring different environments or artistic expressions, while also stating the subject content you want to preserve. The deliverables are edited candidate images for comparison. For tasks that permit changes only to specific areas while all other regions must remain completely unchanged, use clear editing-area controls together with manual review.
If the task is to turn a relatively clear theme into an image, then explore different attributes and styles by modifying the description, DALL·E 2 is worth considering. It is also suitable for generation or editing workflows that already use dall-e-2. When selecting a model, focus on whether candidate images help determine the creative direction; do not treat generated images directly as precise design files.
Compared with DALL·E 2, DALL·E 3 places greater emphasis on understanding and following prompt details, and it also improves results for the same description. If an image contains many interrelated requirements, it is recommended to compare DALL·E 3's generation performance first; if existing images need to be processed, evaluate the editing workflow separately rather than assuming that every operation is more suitable based only on the version number.
Write the subject, environment, visual style, and spatial relationships as a concise description, and avoid cramming an entire design system into the first request.
Choose the image generation endpoint and explicitly provide model=dall-e-2, prompt, and n=1; set size and quality according to this model's documentation, and do not directly reuse shared options from other image models.
Retrieve a URL or Base64 image from the response; save the task ID when asynchronous processing is needed. Check the image details, and if the result includes revised_prompt, save it together with the original requirements.
A fox wearing a blue scarf sits in a wooden boat, with a calm lake in the background, rendered as a watercolor illustration with a simple composition and no text.
Use one clear subject and a small number of attributes, and verify whether the concepts appear together; do not treat shared image parameters as guarantees of arbitrary resolutions or multi-image blending.
No. It accepts image descriptions, or the original image and modification instructions when editing, and primarily delivers images. When you need to discuss ideas or organize requirements first, you can separately complete the text planning, then submit the organized visual description to DALL·E 2; do not expect it to return a complete conversational analysis.
DALL·E 2 is suitable for concept combinations, style exploration, and existing editing workflows; DALL·E 3 places greater emphasis on following details in complex descriptions. If the priority is generating new images with many requirements, you can compare both using the same prompt; modification needs for existing images should be considered separately based on the editing method.
Use /openai/images/edits, provide the original image URI in image and a modification description in prompt, and explicitly specify model=dall-e-2. The description should explain both the intended changes and the content you want to preserve. Text requirements help express intent, but do not strictly lock unchanged areas.
The n parameter for generation and editing endpoints ranges from 1—10, with a default of 1, and can be used to specify the desired number of candidate images. Multiple candidate images are useful for comparing creative directions, but do not mean the content will be completely identical, nor do they guarantee that every image meets all requirements; results should be checked individually before delivery.
You can use response_format to choose url or b64_json, then read the image result from the corresponding field in data. If task_id is returned, you should receive the final result according to the task workflow. Image URLs, Base64 data, and task IDs serve different purposes, and applications need to handle them separately; do not treat a task ID as an image address.
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
↗↗This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs