flux-2-klein is part of Black Forest Labs' FLUX.2 [klein] image series. Designed around a compact architecture and low-latency creation, it combines text-to-image generation with editing of existing images. It is suited to concept exploration, visual concept prototyping, and asset variation production. On this platform, creative tasks can be submitted through a unified image interface, with results received as image links.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
api.acedata.cloudflux-2-kleinInput 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 native positioning of FLUX.2 [klein] is unified generation and editing; the calling specifications here follow this platform's flux-2-klein endpoint.
Use natural language to describe subjects, environments, materials, and lighting to begin text-to-image generation. The klein series emphasizes creative exploration, prompt adherence, and output diversity, making it suitable for comparing different visual directions before refining composition and style. Prompts should highlight key relationships rather than pile up conflicting adjectives.
Generation and editing belong to the same creative workflow: first obtain candidate images, then provide an image link and modification instructions to continue adjusting scenes or visual elements. When editing, clearly specify what should change and what should remain, keeping the task focused on a specific goal instead of redescribing the entire image and causing intent confusion.
Image results are returned as links, making them easy to integrate into asset libraries, preview interfaces, or review workflows. Generation tasks can use count to request multiple candidates; when background processing is needed, asynchronous tasks or callbacks can be used to organize result delivery, allowing users to continue working after submitting an idea instead of staying on a waiting page.
Input character appearances, scene atmospheres, or product concept descriptions to generate visual candidates for team discussion. In each round, adjust only key variables such as composition, materials, or lighting, compare how different directions are expressed, then select images suitable for further refinement; the deliverable is conceptual reference, not a verified physical design.
Provide a product image link, describe the background, environment, and presentation style you want to try, and create contextual visual drafts. Clearly specify in the prompt the appearance, colors, and logos that need to be retained, and review results one by one. Suitable for comparing display concepts during planning; product details must still be verified for accuracy before formal use.
Turn article topics, event atmospheres, or visual requirements for a section into prompts to create illustrations, cover backgrounds, and image candidates. First determine the main subject and intended negative space, then iterate around colors or scenes; titles, dates, and brand text can be added in subsequent layout work to avoid relying on the accuracy of text in generated images.
The klein series emphasizes compact architecture and low-latency creation, making it suitable for work that requires repeatedly trying prompts and comparing visual directions. It should be compared with entry points such as FLUX.2 Pro and Max based on actual results for the same task, rather than judging quality by name alone; if the main goal is the final image, focus on checking details, composition, and preservation after editing.
FLUX.2 [klein] includes different native versions; flux-2-klein is the invocation name used here and should not be directly treated as a version of a particular parameter scale. When choosing, prioritize whether you need text-to-image or editing of an existing image, and whether the result meets the intended use; work requiring multiple reference images or fine-grained reasoning adjustments should use a different entry point with the corresponding inputs and controls.
For text-to-image, specify the subject, materials, lighting, and aspect ratio; for modifying an existing image, prepare image_url and clearly state what to change and retain.
Submit model=flux-2-klein, action=generate, prompt, and size configured according to the API documentation to /flux/images; when editing an image, select edit and add the image URL, while count is used only for generation.
Save the returned image_url; for asynchronous requests, query through task_id or configure callback_url. Compare key elements before and after modifications, then use the completed image in subsequent design work.
Create a clean isometric illustration for a finance app: three coins, a small plant, and a ledger, with teal as the main color, a white background, and no numbers or text.
First check the composition's recognizability, then make minor edits using the selected image; this platform ID is not tied to a specific public parameter scale or fixed speed.
Yes. Use generate for text-to-image, and use edit for modifying existing images, providing an image link and modification prompt. Both task types require prompt and size; edit instructions should clearly describe the intended changes and content to preserve separately, making it easier to verify whether the result meets expectations.
flux-2-klein is this service's invocation name and should not be directly equated with FLUX.2-klein-9B. klein is a model series; when using it, choose based on generation, editing, and output quality. Do not infer parameter count, inference steps, or deployment hardware requirements from this name alone.
The editing input for this endpoint is the string image_url, which is suitable for providing a link to the image to be edited. The native FLUX.2 [klein] series supports multi-reference-image creation, but this does not mean an image array can be placed in image_url; when combining multiple images, use a workflow that explicitly supports the appropriate input method.
After completion, read image_url from the data list in the JSON result to download or display the image. For background processing, use async or callback_url to organize task receipt and associate the task ID; applications should distinguish between successful submission and generation completion, and should not treat task status as an image result prematurely.
Focus prompts on visual requirements such as the subject, composition, and lighting, and avoid making the image carry a large amount of precise text at the same time. After generation, check labels, object relationships, and details that need to be preserved one by one; use post-production layout for titles and descriptions, and finalize important visuals only after comparing multiple candidate rounds.
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
