Wan Image API for scalable text-to-image and editing

The Wan Image API is a unified interface that supports both text-to-image creation and image editing workflows. It is designed for production teams that need repeatable prompts, controllable outputs, and consistent task handling. You can generate new scenes from a prompt or refine existing images using edits, while keeping the same integration surface. Because features can vary by model and region, always confirm capabilities and limits in your EvoLink dashboard and the API responses before you standardize a workflow.

Yes. The Wan Image API includes text-to-image generation and image editing modes. Text-to-image workflows are task-based and support prompt input, while image editing workflows accept reference images and allow fusion or controlled changes. This makes it easy to move from concept generation to refinement without switching tools. For the latest model list and mode availability, check your dashboard and the official API docs, since capabilities can change by region or plan.

The Wan Image API supports prompts in Chinese and English. This is useful for global teams that need to localize creative quickly. For best results, keep prompts structured with subject, style, lighting, and background, and reuse a shared prompt template across regions to maintain consistency. Documented prompt limits include up to 2000 characters for prompts and up to 500 characters for negative prompts when supported, so you have room to be specific without losing clarity.

The Wan Image API image editing mode supports multiple reference images for fusion tasks, which lets you combine elements from different sources. This is useful for merging a product shot with a lifestyle background or blending props into a new scene. The exact count can vary by model, so confirm the current limit in the API docs for your endpoint. Keep one primary image as the anchor and add secondary images only for the elements you want to borrow.

For image editing, supported formats include JPEG, JPG, PNG, BMP, and WEBP. Input images should be between 384 and 5000 pixels on each side and no larger than 10 MB. PNG alpha channels are not supported, so treat PNG inputs as flat images. These constraints help balance output quality with predictable latency. Always verify current limits in your EvoLink dashboard and the API docs because they can evolve with new releases.

The Wan Image API lets you set output resolution using a width x height parameter. For Wan 2.5 image editing, documented total pixels are between 768x768 and 1280x1280, and aspect ratios must be between 1:4 and 4:1. If you do not provide size, the default is 1280x1280. For multi-image input, aspect ratio may follow the last image. Check the API docs for model-specific changes before production deployment.

Yes. Text-to-image generation is task-based in Wan 2.5 and earlier, and image editing also runs as tasks. You submit a request, receive a task id, then poll the task endpoint until it completes. Task ids are documented as valid for 24 hours, so retrieve results promptly and store them in your asset pipeline. In production, use backoff for polling to avoid unnecessary retries while keeping acceptable latency.

The Wan Image API allows multiple outputs per request, which is useful for A/B testing and creative review. For Wan 2.5 image editing, the documented output count parameter n can be between 1 and 4, and the number of images impacts cost directly. A practical workflow is to generate three or four candidates in the first pass, choose the strongest, and then run smaller refinement batches to control spend while keeping quality high.