LogoWan API Docs
Homepage
X (Twitter)
Homepage

Get started

OverviewIntegration guideCheck balanceDelete a task

Wan models

Wan 3.0Wan 3.0 PrimeWan 3.0 ProWan 3.0 Prime Pro

Seedance models

Seedance 2.5Seedance 2.5 FlashSeedance 2.0Seedance 2.0 FastSeedance 2.0 Mini

Image models

Seedream 5.0 ProSeedream 5.5 API

Language models

Claude APIClaude Haiku 5.5Claude Sonnet 5.5Claude Opus 5.5Claude Fable 5.2GPT APIGPT Luna 6GPT Terra 6.1GPT Sol 6.1GPT Astra 6
X (Twitter)

Seedream 5.5 API

Generate and edit images or split layers with Seedream 5.5. Request parameters, asynchronous tasks, image metadata and current per-image pricing.

Seedream 5.5 generates one image from text or up to 10 reference images. Layer decomposition takes one image and returns a base image plus transparent layers, with an optional printed-text recognition result.

Try the Playground · OpenAPI JSON · Integration tutorial

Authentication and endpoints

Use your Wan API key as a Bearer token. One key accesses all available models and groups; you do not select a group on the key.

MethodURLPurpose
POSThttps://api.wanapi.app/seedream55/v1/images/generationsSubmit one image task
GEThttps://api.wanapi.app/seedream55/v1/tasks/{id}Read that task

The model ID is seedream-5-5. These are asynchronous endpoints. Submit once, save the task ID and poll it until completed or failed. An uncertain network response is not permission to submit again.

Generate an image

curl https://api.wanapi.app/seedream55/v1/images/generations \
  -H "Authorization: Bearer $WAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedream-5-5",
    "prompt": "A ceramic teapot shaped like a sleeping fox on a linen tablecloth, soft morning light. No text.",
    "size": "1K"
  }'

An accepted request returns a task rather than image bytes:

{"id":"task_example","model":"seedream-5-5","status":"processing","created_at":1794432000}
curl https://api.wanapi.app/seedream55/v1/tasks/task_example \
  -H "Authorization: Bearer $WAN_API_KEY"

Poll the same ID every few seconds. Keep it if your application stops waiting, so it can resume later. Task IDs and timestamps in this guide are illustrative.

Parameters

ParameterTypeDefaultRules
modelstringrequiredseedream-5-5
promptstringrequired for generationOptional for layer decomposition.
imageURL or URL arraynoneUp to 10 references; exactly one for layers. Every reference is billed.
sizestring2K, or auto for layersSee the size rules below.
output_formatstringpngpng or jpeg. For layers, controls only the base image.
backgroundstringautoauto, opaque, transparent. Transparent requires PNG. Omit this field entirely in layer mode.
layer_decompositionbooleanfalseSplit one reference into a base image and transparent PNG layers.
layer_decomposition_options.max_imagesinteger30Total image cap, 2–100, including the base. Only for layer mode.
layer_decomposition_options.text_recognitionbooleanfalseRequest printed-text metadata in layers. Only for layer mode.
optimize_prompt_options.modestringstandardThe only supported mode is standard.
watermarkbooleanfalseRequest an AI-generated watermark.
content_filterbooleantrueThe customer API accepts false. In the Playground, only administrators may disable it.

There is no supported n, seed, stream, resolution, aspect_ratio, image_urls or response_format parameter. Undocumented fields are discarded and do not enable those features. Multipart uploads, Base64 and data URIs are rejected.

Output size

Generation accepts 1K, 1.5K, 2K or exact WIDTHxHEIGHT pixels. Exact sizes must total 921,600–4,624,220 pixels and have an aspect ratio from 1:16 to 16:1. For example, 2048x1152 is valid. A ratio such as 16:9 is not a size value.

Layer decomposition accepts auto, 1K, 1.5K or 2K; exact pixels are not supported in that mode. auto follows the input and reserves the higher output rate. The model has no 3K or 4K preset.

Reference images and editing

Provide public HTTP(S) URLs hosted by you, without embedded credentials. Supported formats are JPEG, PNG, WebP, BMP, TIFF, GIF, HEIC and HEIF, up to 30 MB per image and at most 36,000,000 pixels in total. AVIF and JP2 are unsupported. A URL must serve an image directly without a login or browser challenge.

{
  "model": "seedream-5-5",
  "prompt": "Change the scene to a snowy winter night. Keep the lighthouse and its composition.",
  "image": "https://your-cdn.example/lighthouse.jpg",
  "size": "2K"
}

For multiple references, pass a URL array and explain each image's role in the prompt. The result is still one image. Reference URLs in examples are placeholders: replace them with your own reachable files.

Layer decomposition and text recognition

{
  "model": "seedream-5-5",
  "image": "https://your-cdn.example/poster.png",
  "size": "1K",
  "layer_decomposition": true,
  "layer_decomposition_options": {
    "max_images": 4,
    "text_recognition": true
  }
}

The cap includes the base image; 4 means at most four returned images, not four layers plus a base. The model may return fewer. Leave prompt out for automatic separation, or name the elements to separate. A prompt can suppress text that it excludes.

Layers are PNG files with transparency. Rebuild the image in ascending z_index order, with the base at 0. If a bounding box is provided, resize the layer to that box before positioning it: the layer file's pixel dimensions can differ from its placement dimensions.

Optional text_recognition contains full text and lines. Each line can include a bounding_box and styled runs. A run's style can include similar_font_family, foreground_color (#RRGGBB), bold, italic, underline, strikethrough, shadow and translucent. These are model estimates. Artistic or handwritten lettering may have no text result; treat the field as optional.

Completed response

{
  "id": "task_example",
  "model": "seedream-5-5",
  "status": "completed",
  "created_at": 1794432000,
  "output": {
    "data": [
      {
        "url": "https://api.wanapi.app/v1/tasks/task_example/artifacts/image-1/content?access=EXAMPLE",
        "size": "1248x832",
        "output_format": "png"
      }
    ]
  }
}

Each image has a url and size. Other fields can include output_format, background, z_index, name, description, bounding_box and text_recognition. bounding_box.absolute is [left, top, right, bottom] in base-image pixels. bounding_box.normalized uses the same coordinates scaled to 0–1000. Do not assume optional metadata exists on every image.

Treat image URLs as private download links and save the files you need. A URL carrying an access token can be opened without exposing your account API key.

Pricing

Billable imageUSD per image
Output up to 2.61 MP (1K / 1.5K)$0.017
Output above 2.61 MP (2K)$0.02975
Every reference image, including the first$0.0068

Rates refresh automatically. One API key works across all models and groups.

The output pixel boundary is 2,610,000 pixels. Calculate the bill as:

lower-band output count × lower-band rate
+ higher-band output count × higher-band rate
+ reference image count × reference rate

Every reference is billed, including the first and the source image for layers. Generation and layers use the same output rates. Submission reserves one output, or the layer max_images cap, plus references. Completion uses each returned image's own pixel dimensions and releases the unused reservation. The accepted task keeps its price snapshot even if rates change while it runs. Failed tasks are refunded.

Errors and retry behavior

Invalid parameters are rejected before submission. Insufficient balance or key quota prevents generation. Authentication and ownership checks apply to both submission and polling. A failed task returns status: "failed" with an error object.

Retry a failed status read with the same task ID. Do not automatically repeat a generation POST, including after a timeout: it may already have been accepted. While the upstream reports normal progress, the task remains pending. A task with no normal upstream status or successful delivery for three continuous hours fails and is refunded.

See the Seedream 5.5 tutorial for an application loop that saves and resumes the task ID.

Seedream 5.0 Pro

Seedream 5.0 Pro API reference: text-to-image, reference editing, layer decomposition, transparent PNG, request fields, responses and pricing.

Claude API

Call Claude through Anthropic Messages or OpenAI Chat Completions. Model IDs, streaming examples and live input, output and cache prices by group.

Table of Contents

Authentication and endpointsGenerate an imageParametersOutput sizeReference images and editingLayer decomposition and text recognitionCompleted responsePricingErrors and retry behavior