Wan 2.5 Image Editing Preview

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit

Not collected

Maximum input tokens

Not collected

Maximum output

Not collected

Input modalities

Text, ImageSource for Input modalities (opens in a new tab)

Output modalities

ImageSource for Output modalities (opens in a new tab)

Reasoning

Not collected

Tool calling

Not supportedSource for Tool calling (opens in a new tab)

Singapore card scope. See individual regions for differences.

Structured output

Not supportedSource for Structured output (opens in a new tab)

Singapore card scope. See individual regions for differences.

Streaming

Not collected

Batch processing

Not supportedSource for Batch processing (opens in a new tab)

Singapore card scope. See individual regions for differences.

Fine-tuning

Not supportedSource for Fine-tuning (opens in a new tab)

Singapore card scope. See individual regions for differences.

Knowledge cutoff

Not collected

Release date

2025-09-23Source for Release date (opens in a new tab)

Earliest reviewed regional release, not inferred from selector version.

Parameters

Not collected

Architecture

Not collected

Languages

Chinese, EnglishSource for Languages (opens in a new tab)

Documented prompt languages.

Context caching

Not supportedSource for Context caching (opens in a new tab)

Singapore card scope. See individual regions for differences.

Search grounding

Not supportedSource for Search grounding (opens in a new tab)

Singapore card scope. See individual regions for differences.

Prefix completion

Not supportedSource for Prefix completion (opens in a new tab)

Singapore card scope. See individual regions for differences.

Console model playground

SupportedSource for Console model playground (opens in a new tab)

API aliases

wan2.5-i2i-previewSource for API aliases (opens in a new tab)

The repeated snapshot section uses the same selector and does not establish another model.

Maximum prompt characters

2,000Source for Maximum prompt characters (opens in a new tab)

Characters, not tokens. Excess prompt characters are truncated.

Output formats

PNGSource for Output formats (opens in a new tab)

Visible watermark

Optional AI Generated watermark, off by defaultSource for Visible watermark (opens in a new tab)

Random seed range

0–2,147,483,647; for multiple images, successive seeds are assigned from the starting seedSource for Random seed range (opens in a new tab)

The same seed does not guarantee identical output.

Output link expiry

Task IDs and generated image URLs expire after 24 hoursSource for Output link expiry (opens in a new tab)

This does not establish a complete input retention or training-use policy.

Output billing

Only successful output images count toward usage, including partial-success responses. Failed images do not add to image_count.Source for Output billing (opens in a new tab)

Usage metering

Price × actual generated image count, not text tokensSource for Usage metering (opens in a new tab)

Asynchronous generation

SupportedSource for Asynchronous generation (opens in a new tab)

Asynchronous jobs use X-DashScope-Async: enable and task polling. Sync and async are transports of the same model.

Request mode

Asynchronous legacy image2image API; X-DashScope-Async: enable, then poll the task ID. The newer messages/content protocol is not used.Source for Request mode (opens in a new tab)

Required request fields

input.prompt and input.images are required; image order mattersSource for Required request fields (opens in a new tab)

Negative prompts

At most 500 characters; excess is truncatedSource for Negative prompts (opens in a new tab)

Images per request

n=1–4 generated images; default 4Source for Images per request (opens in a new tab)

Aspect ratios

1:4 to 4:1Source for Aspect ratios (opens in a new tab)

Prompt rewriting

prompt_extend=true by default; optional LLM-based prompt rewriting adds processing time.Source for Prompt rewriting (opens in a new tab)

Resolution controls

Explicit width*height; when omitted, approximately 1280×1280 total pixels following the last input image aspect ratio.Source for Resolution controls (opens in a new tab)

Output resolutions

768×768 to 1280×1280 total pixelsSource for Output resolutions (opens in a new tab)

Image-to-image editing

SupportedSource for Image-to-image editing (opens in a new tab)

Reference images

3Source for Reference images (opens in a new tab)

Input image limits

One to three ordered reference images; each up to 10 MB, 384–5,000 pixels per side. Public HTTP(S) URL or Base64.Source for Input image limits (opens in a new tab)

Input image formats

JPEG, JPG, PNG (no alpha), BMP, WEBPSource for Input image formats (opens in a new tab)

Published availability regions

China (Beijing) · Chinese Mainland, Singapore · InternationalSource for Published availability regions (opens in a new tab)

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

Why it is included

Classification inference: a separately documented native image API release has no exact checkpoint in the reviewed official Wan-AI public-weight inventory. Public Wan video checkpoints are not evidence of the image model weights. Private lineage, parameters and architecture remain unknown.

Lifecycle dates

Sources reviewed