Alibaba Cloud's Wan3.0 Video Generation Model Officially Launches, Generating 30-Second Videos in a Single Run and Supporting Document Input

en.Wedoany.com Reported - On August 24, Alibaba Cloud announced the official launch of its video generation model, Wan3.0. The model began public beta testing on August 6, featuring comprehensive upgrades in generation duration, input modalities, and content understanding. It can generate high-definition videos of up to 30 seconds in a single run and, for the first time, supports direct input of document formats such as doc, xls, ppt, pdf, and md for video generation.

Wan3.0 is an all-in-one reference video generation foundation model launched by Alibaba Cloud, belonging to the Tongyi Wanxiang (Wan) series, with inference services provided by the Alibaba Cloud Bailian platform. Compared to its predecessor, the Wan 2.7 series, which supports a maximum generation duration of 15 seconds, Wan3.0 doubles this capability. The model deeply integrates multimodal input capabilities including text, images, videos, and audio, uniformly supporting multiple generation modes such as text-to-video, image-to-video (first frame/first and last frames), and reference-based video generation. Users can upload up to 10 images, 5 videos, and 5 audio clips as multimodal reference materials; for file input, a maximum of 1 file or link can be uploaded per session, with a file size limit of 100MB and no more than 50 pages. The model supports output resolutions of 480P, 720P, and 1080P, with aspect ratios adaptable to various formats including 16:9, 4:3, 1:1, 3:4, and 9:16.

In terms of physical detail restoration, Wan3.0 delivers improved rendering of subtle features such as hair, glass reflections, and water surface gloss, reducing the likelihood of character distortion and visual glitches. In all-in-one reference mode, the model can precisely replicate character appearances, prop details, spatial relationships, and lighting styles, ensuring high consistency of key visual elements during long-video or multi-shot generation. During the public beta phase, Wan3.0 has been applied across multiple industries: in short-drama and film production, it can produce close-up facial shots, improving the usability rate for large-scale output; in the advertising industry, it stably preserves product appearance and logo details, with results approaching real-shot quality; in cultural tourism scenarios, it can quickly generate promotional videos for city landmarks; and in the music sector, it supports image input to automatically generate multi-shot music videos with synchronized lyrics and lip movements.

In terms of pricing, Wan3.0 is billed on a pay-as-you-go basis, with API prices for 480P, 720P, and 1080P set at 0.3 yuan/second, 0.6 yuan/second, and 1.2 yuan/second, respectively. From August 24 to September 23, APIs on the Alibaba Cloud Bailian and Qwen AI platforms are eligible for a limited-time 30% discount. Currently, users can experience the model on platforms such as Alibaba Cloud Bailian, the Wanxiang official website, Wanjing Yike, the Qwen AI platform, Qwen Creation PC client, the Qwen App, Duiyou, and IF STUDIO. On the ecosystem front, Maiya's Tiaoyue Shijie, Deevid, Juhuo, JD Lingjing Platform, Meitu, and Jingmeng have already integrated Wan3.0; platforms including Meitu, JD Lingjing, iQIYI Nado Pro, and Bilibili Updream have also completed integration.