Ali Wan 3.0, another step forward for productivity tools
On August 6, Alibaba’s new generation video generation model Wan3.0 began public beta testing. It can generate videos up to 30 seconds in length per time. This upgrade covers three aspects: generation duration, versatile creativity, and realism: it supports input from structured documents such as doc, xls, and ppt; it maintains stability across shots in terms of characters, props, and spatial relationships; and it improves the realistic reproduction of characters’ personalities. The Wan3.0 API is priced at 0.3 to 1.2 yuan per second, and all interfaces have been recently made available. Compared to previous fragmented generation methods, Wan3.0 can fully express complex shot languages and understand the semantics of office materials, filling the gaps in the video model’s input端; it optimizes skin details and micro-expressions for character portrayal, eliminating the problem of “standard faces”; and it ensures consistency in long videos, preventing deviations in characters, sound fields, and style during multi-shot transitions. This model has already been tested in commercial applications such as film, advertising, design, and cultural tourism, facilitating the transition of AI video from a toy to a productivity tool.