Models / WAN / WAN 3
WAN 3 AI Video Generator
WAN 3 by Alibaba is an AI video model, released in August 2026. Generate with WAN 3 on AgentDocks: text to video with native audio, 2 to 30 seconds, 480p to 1080p. Priced per second in credits.
Overview
What is WAN 3?
WAN 3 is a version of WAN, Alibaba's family of AI video models, released in August 2026. Text, image or reference to video with native audio. Two to thirty seconds, 480p to 1080p, six aspect ratios including adaptive.
Thirty Seconds in One Pass
Generate any length from two to thirty seconds with sound, without stitching clips together.
Start and End Frames
Give WAN 3 a first still and an optional last one, and the clip moves between them.
References by Number
Attach up to ten reference images and address each one in the prompt by its position.
What Changed
What WAN 3 Adds
01
Native audio generated with the clip, where 2.6 was silent.
02
Any whole-second length from two to thirty, where 2.6 offered five, ten or fifteen.
03
A 480p tier for cheap drafts.
Made on AgentDocks
What WAN 3 Does Best
Sample work made on AgentDocks, showing the kind of results WAN 3 is built for.
Holds a Scene for Thirty Seconds
Generates Sound with the Picture
Builds Shots From References
Also on the Canvas
WAN is One Menu Entry. Here are the Others.
