Nano Banana

Wan 3.0 — Alibaba's AI Video Generator with Native Sound and Realistic Faces

Wan 3.0 is Alibaba's latest video model. It generates natural faces and expressions, keeps characters and props consistent, and adds dialogue, music and sound effects. Type a prompt below to try it.

Open the full workspace

More models

Why Wan 3.0 Is a Great AI Video Generator

Wan 3.0 is a flexible, affordable video model. It is strong on people — faces, micro-expressions and speech — and it can build scenes from as many as 10 reference images.

  • Realistic faces

    Natural skin, eye movement and micro-expressions make close-ups and talking shots convincing.

  • Dialogue, music and effects

    Generates spoken lines, background music and sound effects together with the picture.

  • Consistent characters and props

    Use up to 10 reference images to keep people, products and objects looking the same.

  • Stable text and UI

    On-screen text, signs and interfaces stay readable instead of warping.

What Wan 3.0 can do

  • Text to video

    Describe the scene, the action and the sound in one prompt.

  • First and last frame

    Upload 1 image to start from, or 2 to set the start and end of the clip.

  • Up to 10 reference images

    Combine characters, outfits, products and locations from several images into one scene.

  • Flexible durations

    2, 3, 4, 5, 10 or 15 seconds — from a quick loop to a full short scene.

  • Affordable drafts

    Start at 480p for low-cost tests, then render the final at 720p or 1080p.

  • 6 aspect ratios

    16:9, 4:3, 1:1, 3:4 and 9:16, or Adaptive to follow your input image.

Use cases

Starting points you can copy into the generator above.

Use caseGood forExample prompt
Talking-head clipsPresenters and characters speaking with natural expressions.

A friendly woman in her thirties sitting in a bright home office looks at the camera and says "Here are three tips for better sleep", gentle smile, natural light, soft background music

Product showcasesKeep a product exactly on-brand using reference images.

The headphones from the reference images rest on a wooden desk, a hand picks them up and puts them on, the screen behind shows the text "Focus Mode On", subtle electronic sound

App and UI demosScreens and on-screen text that stay legible.

Close-up of a phone screen scrolling through a recipe app, the title "Easy Pasta Night" clearly visible, a finger taps the Start button, soft click sounds

Character seriesThe same character across several clips.

The boy in the yellow hoodie from the reference images skateboards down a seaside promenade at sunset, waves and wheels on wood, tracking shot

Quick loops2–3 second clips for stickers, banners and backgrounds.

A steaming cup of tea on a windowsill while rain runs down the glass, seamless calm loop, soft rain sound

Music-backed scenesShort mood clips with generated background music.

A night drive through a city with neon reflections on the windshield, lo-fi hip hop music playing, slow dreamy camera, 16:9

Wan 3.0 in 3 steps

  1. 1

    Describe the scene or add images

    Write the action, lines and sounds, or use first/last frame (1–2 images) or reference mode (1–10 images).

  2. 2

    Pick duration, resolution and format

    Choose 2–15 seconds, 480p to 1080p, an aspect ratio and whether to generate audio. The credit cost is shown before you submit.

  3. 3

    Generate and keep your video

    Videos take a few minutes and are saved to My works. If a video fails, its credits are refunded automatically.

X Posts About Wan 3.0 — Alibaba's AI Video Generator with Native Sound and Realistic Faces

YouTube Videos About Wan 3.0 — Alibaba's AI Video Generator with Native Sound and Realistic Faces

Frequently asked questions

People and consistency: realistic faces and expressions, characters and props that stay the same across shots, readable on-screen text, plus native dialogue, music and sound effects.

Ready to Mix Every Input with Wan 3.0?

Multimodal video from text, frames or up to 10 references, in 1080p with native audio.

Explore All Models
  • Clips up to 15s
  • Native audio
  • Failed outputs refunded