flux-kontext-pro is the image generation and editing model for Black Forest Labs FLUX.1 Kontext [pro], focused on context-aware modifications to existing images. It can create new images from text and perform editing instructions based on an original image, making it suitable for adjusting product displays, character assets, and design drafts. On this platform, generation, editing, and result retrieval can all be completed through the same image API.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
api.acedata.cloudflux-kontext-proInput 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 above are this platform's API specifications; the native capabilities of FLUX.1 Kontext [pro] and the control options available through the API should be understood separately.
When editing, there is no need to describe the entire image from scratch. Instead, provide the original image and specify the object, attribute, or area you want to change. For example, adjust an object's color, add accessories to a person, or modify a specified element in the image. Its focus is on making changes based on image context while maintaining consistency with the original composition and unmodified content.
The same model provides both text-to-image generation and image editing. You can first create a draft using descriptions of the scene, subject, lighting, and aspect ratio, then use the selected result as editing input to continue making modification requests. This workflow is suitable for first establishing a visual direction and then progressively refining details, rather than regenerating the entire image each time.
Results are delivered as image links, making them convenient for business systems to display, download, or send into subsequent review workflows. When backend processing is needed, you can use asynchronous task queries or configure completion callbacks. Applications can associate task identifiers with asset records, separating submission, waiting, and result reception without requiring users to remain on the submission page.
Provide an existing product showcase image, specify the colors, scene elements, or decorations that need adjustment, and clearly state the product subject and composition you want to preserve. The output can be used to compare different presentation options and help designers select a direction. When packaging text, trademarks, and product structure are involved, check each item before adoption to avoid mistaking generated changes for physical product details.
Based on existing photos or character design images, request specific changes such as adding accessories or adjusting clothing elements to obtain modified candidate assets. Compared with describing a character from scratch and generating it again, the original image provides clear visual context. Suitable for avatar design, event visuals, and character draft adjustments, but facial features and key identifying characteristics still require manual review.
First generate a concept image, then successively modify objects, color schemes, or scene details in the selected draft to create options that are easy to review. Save the input image, instructions, and results at each step to clearly show the modification process. The output is suitable as a visual proposal and reference for subsequent production, rather than directly replacing design files that require precise layers, dimensions, or structure.
When the task is to preserve the main content of an image while changing only some elements, Kontext Pro's context editing is better suited to the need. If you simply want to explore entirely new compositions from text, you can also use its generation mode; distinguish between “creating a new image” and “modifying an existing image” when choosing, and compare results using actual assets rather than relying only on model names.
Kontext Pro is suitable as a starting point for everyday context editing; when facing complex editing requirements, include Kontext Max in the comparison and evaluate detail performance using the same original image and instructions. Kontext [dev] is another open-weight version; do not directly apply its local deployment methods, parameters, or license to Pro, and do not treat FLUX 2's capabilities as features of this model.
For text-to-image, clearly specify the subject, materials, lighting, and aspect ratio; for modifying an existing image, prepare image_url and clearly describe what to change and preserve.
Submit model=flux-kontext-pro, action=generate, prompt, and size=16:9 to /flux/images; when editing images, select edit and provide the image URL, while count is only used for generation.
Save the returned image_url; for asynchronous tasks, query using task_id or configure callback_url. Compare key elements before and after modification, then use the finished image in subsequent design work.
Keep the chair outline, camera angle, and background from the original image, changing only the seat color from gray to dark green while keeping the shadows and ambient light harmonious.
Compare against the original image to inspect the chair legs, edges, and background; save the result after each round, and in the next round request only additional changes rather than treating editing as lossless cutout extraction.
Yes. Use the generate operation, submit a prompt and image aspect ratio size, and explicitly specify model as flux-kontext-pro. The prompt can describe the subject, scene, composition, and lighting; if the goal is to modify existing material, use edit instead and provide the original image link.
First identify the object to modify, then describe the desired change, and add what needs to be preserved. For example, “Change the bottle body to dark blue, while keeping the cap, label, and background composition.” Try to avoid requesting multiple vague changes at once; complex tasks can be split into consecutive steps, checking the result after each step before continuing.
This model uses image aspect ratios to control the canvas, so enter ratios such as 1:1, 16:9, or 9:16 in size rather than pixel dimensions. Ratios are used to express square, landscape, or portrait compositions, and should not be understood as a commitment to a fixed output pixel size.
The editing input here is the original image specified by image_url, and count is not used for editing tasks. When processing multiple assets, submit editing tasks separately and manage their results individually. Generation tasks can use count to request multiple candidates, but this is not the same feature as multi-original-image editing or multi-image reference.
You can use the image link returned from the previous round as the image_url for the next round, then submit a new edit instruction to make step-by-step changes. It is recommended to keep the original image and the results from each round for easy rollback and comparison; with consecutive edits, you should still check changes to the subject, text, and composition, and not assume that all other content will be completely preserved in every round.
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
