Wan 3.0, Alibaba's new video generator with artificial intelligence , allows you to combine up to 20 references in the same project: images, videos, audios, documents and even web pages. The tool promises to preserve the details of these materials more accurately during generation.

In addition, it generates immersive audio with music and sound effects in an integrated way, and produces clips of up to 30 seconds. Its presentation adds another competitor in the race for the creation of audiovisual content with AI, where models such as Sora, Veo and Seedance also compete for greater duration and quality.

Announced on August 6, 2026 as a public beta, Wan 3.0 unifies in a single model capabilities that Wan 2.7 required from four separate models. Additionally, it is the first time that a video model accepts office documents (PDF, DOC, XLS, PPT) as native input, allowing presentations and reports to be transformed directly into videos.

Unlike Wan 2.1 and 2.2 (open source under Apache 2.0), Wan 3.0 is API-only, with no public weights. The maximum confirmed resolution is 1080p (4K rumors were officially dismissed), and the cost is around $0.05 to $0.20 per second, which implies approximately $6 for a 30-second clip. Access is limited to guest accounts on Alibaba Cloud Model Studio; the Beijing region offers prices 20% cheaper than Singapore.

## Sources

https://morphic.com/resources/models/wan-3-0

https://www.ngram.com/blog/wan-3-0-document-to-video-ai-model

https://www.gptunnel.ru/en/blog/wan-3-0-news

https://www.toolworthy.ai/tool/wan-3-0

https://zencreator.pro/ai-university/guides/wan-ai-video-model-guide