Wan 2.6 Reference to Video

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, Image, VideoSource for Input modalities (opens in a new tab)

API parameter contract; regional card modality labels can be incomplete or inconsistent.

Output modalities

Video, AudioSource for Output modalities (opens in a new tab)

Audio is synchronized within the generated video, not a separate TTS endpoint.

Reasoning

Not collected

Tool calling

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

Singapore card scope; individual regional cards are preserved under access.

Structured output

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

Singapore card scope; individual regional cards are preserved under access.

Streaming

Not collected

Batch processing

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

Singapore card scope; individual regional cards are preserved under access.

Fine-tuning

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

Singapore card scope; individual regional cards are preserved under access.

Knowledge cutoff

Not collected

Release date

2025-12-16Source for Release date (opens in a new tab)

Earliest reviewed platform release across regions.

Parameters

Not collected

Architecture

Not collected

Languages

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

Prompt languages, not a full generated-speech language list.

Context caching

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

Singapore card scope; individual regional cards are preserved under access.

Search grounding

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

Singapore card scope; individual regional cards are preserved under access.

Prefix completion

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

Singapore card scope; individual regional cards are preserved under access.

Console model playground

Not supportedSource for Console model playground (opens in a new tab)

API aliases

wan2.6-r2vSource for API aliases (opens in a new tab)

The duplicated Wan 2.5 snapshot block repeats the same selector and is not an additional model.

Maximum prompt characters

1,500Source for Maximum prompt characters (opens in a new tab)

Chinese characters or English letters count as characters; excess input is truncated. This is not a token limit.

Negative prompts

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

Frame rate

30Source for Frame rate (opens in a new tab)

Output formats

MP4 (H.264)Source for Output formats (opens in a new tab)

Video duration

2–10 secondsSource for Video duration (opens in a new tab)

Default duration is 5 seconds.

Output resolutions

720P, 1080PSource for Output resolutions (opens in a new tab)

Default resolution

1080PSource for Default resolution (opens in a new tab)

Videos per request

1Source for Videos per request (opens in a new tab)

Visible watermark

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

Random seed range

0–2,147,483,647; omitted selects a random seedSource for Random seed range (opens in a new tab)

Same seed does not guarantee identical output.

Request mode

Legacy DashScope video-synthesis protocol; submit with X-DashScope-Async: enable, then poll /api/v1/tasks/{task_id}Source for Request mode (opens in a new tab)

Task ID lifetime

24 hours for task queriesSource for Task ID lifetime (opens in a new tab)

Output link expiry

Result URL expires after 24 hours; save generated video promptlySource for Output link expiry (opens in a new tab)

Required request fields

prompt plus reference_urls; one to five references, with at most three videosSource for Required request fields (opens in a new tab)

Reference images

Up to five images; images plus videos may not exceed fiveSource for Reference images (opens in a new tab)

Image input limits

JPEG/JPG/PNG (no transparency)/BMP/WEBP; 240–8,000 pixels per side; at most 20 MB; public URLSource for Image input limits (opens in a new tab)

Source video limits

At most three MP4/MOV reference clips; 1–30 seconds and at most 100 MB each; public URLSource for Source video limits (opens in a new tab)

Reference image syntax

character1, character2, etc. follow reference_urls order; each subject reference should contain one character.Source for Reference image syntax (opens in a new tab)

Legacy API aliases

reference_video_urls is deprecated: up to three videos, 2–30 seconds each. The replacement reference_urls also accepts images and one-second video inputs.Source for Legacy API aliases (opens in a new tab)

Generation controls

shot_type single/default or multi overrides the prompt; no prompt_extend dependency is stated in this APISource for Generation controls (opens in a new tab)

Resolution controls

size must be width*height; defaults to 1920*1080Source for Resolution controls (opens in a new tab)

Aspect ratios

16:9, 9:16, 1:1, 4:3, 3:4Source for Aspect ratios (opens in a new tab)

Output resolutions

1280×720, 720×1280, 960×960, 1088×832, 832×1088, 1920×1080, 1080×1920, 1440×1440, 1632×1248, 1248×1632Source for Output resolutions (opens in a new tab)

Source audio handling

Uses reference-character voice when available and generates synchronized audio; audio=false is documented only for Flash.Source for Source audio handling (opens in a new tab)

Output billing

Output seconds plus sum min(video duration, 5 / total reference count). Both images and videos count in the divisor; images themselves are free. Total billable input is at most 5 seconds.Source for Output billing (opens in a new tab)

Reference video billing

With one video and four images, billable input is at most one second. The guide gives 1.65 for 5/3 in examples without defining general rounding precision.Source for Reference video billing (opens in a new tab)

Usage metering

Billed seconds = capped reference input + outputSource for Usage metering (opens in a new tab)

Published availability regions

Singapore · International, China (Beijing) · Chinese Mainland, Germany (Frankfurt) · Global, US (Virginia) · GlobalSource 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: this native Wan 2.5/2.6 API release is separately documented and has no exact checkpoint in the reviewed official Wan-AI public-weight inventory. Public Wan 2.1/2.2 checkpoints are excluded. API availability and inventory absence do not prove private checkpoint lineage; undisclosed parameter counts and architecture are not borrowed from public models.

Lifecycle dates

Sources reviewed