Skip to content
Wan3.0
Public beta

Turn anything into video

Wan 3.0 generates up to 30 seconds in a single pass and reads documents, spreadsheets and decks as source material — not just text and images.

Wan 3.0 output
Single-pass duration
30sSingle-pass duration
Maximum resolution
1080pMaximum resolution
Document formats
9Document formats
Max file size
100MBMax file size
Max pages
50Max pages
Input types
5Input types

Duration

Thirty seconds, one continuous take

Longer generation is not just a bigger number. Thirty seconds gives narrative room — continuous camera movement, one-take shots, and complex camera language that a short clip cannot hold. It moves the model from generating a shot to telling a complete story.

Smart Duration

Recommends an appropriate length automatically, based on your prompt.

Video Extension

Extend an existing result to develop the story and let the plot grow further.

30s, single pass

A single shot

A complete story

Universal input

Anything can become video

Beyond text, image, audio and video, Wan 3.0 is the first in the family to read documents directly. Office material becomes video without an intermediate step.

Omni-creation

Five input types

  • Text

  • Image

  • Audio

  • Video

  • Document

    New

Supported document formats

  • .doc
  • .xls
  • .ppt
  • .pdf
  • .txt
  • .key
  • .pages
  • .numbers
  • .md

Web links are supported as input too.

Input limits

  • One file or link per generation
  • 100MB maximum file size
  • 50 pages maximum

What you can produce

  • Teaching courseware

    Turn lesson decks and notes into a paced explainer.

  • Product demos

    A product deck can provide source material for a presentation video.

  • Animated charts

    Spreadsheet content can guide animated charts.

  • Business reports

    Reporting documents become something people will watch.

Fidelity

Facial detail with restrained expression

Wan 3.0 renders facial features and skin detail with restrained, natural expression and micro-expressions coordinated with body language.

Reference consistency

Consistency where production needs it

Across omni-reference tasks, Wan 3.0 keeps reference details consistent.

  • 01

    Characters

    Facial features, hairstyle and colour, physique, clothing and accessories stay stable — so a character does not drift out of role.

  • 02

    Props

    Multi-angle appearance, hardware structure, logos and material detail remain consistent.

  • 03

    Scenes

    The relationship between character blocking and camera position is handled correctly.

  • 04

    Style

    Cinematic tonality is rendered precisely, and multi-style work resists style bleed.

Charts, data and interfaces

For digital and graphic material, Wan 3.0 raises the aesthetic baseline — harmonious colour and fluent delivery, so charts, data and UI carry the same quality as filmed content.

Editing

Video editing is substantially improved. You can revise the visuals, the plot, and the dialogue.

  • Visuals
  • Plot
  • Dialogue
Instruction-based editing

Immersive experience

Realism you can hear as well as see

Wan 3.0 pushes on realism, visual texture and sound design together, for a stronger audiovisual impact than picture alone can carry.

With sound
Press play and unmute to hear the sound design.
  • Realism

    Light, motion and material behave the way the eye expects.

  • Visual texture

    Surface detail holds up under attention instead of flattening out.

  • Sound design

    Audio is composed with the image, not laid over it afterwards.

Known limitations

What still needs work

Alibaba Cloud states plainly that audio texture and text accuracy still have room to improve. If your work depends on precise on-screen typography or high-fidelity sound, plan for a pass in post.

Lineage

Eight versions, from Wan 1.0 to Wan 3.0

The Wan family has iterated through eight releases. Wan 3.0's upgrades come from real industry requirements, and are built to return to real production pipelines.

Wan 1.08versionsWan 3.0

In production

Built for real creative pipelines

  • Film & episodic

    Film & episodic

    Thirty-second duration and reference consistency support AI film, short drama, music and dance video, and vlog workflows.

    AI film · Short drama · Music MV · Vlog

  • Advertising & marketing

    Advertising & marketing

    Text, image and other supported source materials can guide customised advertising and marketing video.

    Appliances · Cosmetics · Automotive · FMCG · 3C · Apparel · Software

  • Design & creative

    Design & creative

    Product UI walkthroughs, software feature animation and data visualisation can use interfaces, documents and charts as source material.

    UI demos · Feature animation · Data visualisation

  • Travel & culture

    Travel & culture

    Reference material about scenery, cultural landmarks and local food can guide city films and cultural digitisation workflows.

    City films · Landmarks · Cuisine · Cultural archives

FAQ

Common questions

Wan 3.0 generates up to 30 seconds in a single pass. Smart Duration can recommend a length from your prompt, and Video Extension can extend a result further.

Start with a prompt, a deck, or a link

Wan 3.0 is in public beta. Bring the material you already have.