Alibaba Cloud Unveils Wan3.0 Model that Generates 30-Second Videos

Summary

Alibaba Cloud has officially launched Wan3.0. This video generation model can create clips of up to 30 seconds and accepts structured document inputs (DOC, XLS, PPT, PDF, Markdown). The model has been in public beta since early August, and a 30% API discount valid until September 23 is still in effect.

Why it matters

The 30-second generation duration represents a significant step toward usable narrative content rather than isolated short clips. The ability to directly ingest office documents as creative input indicates that Alibaba is targeting integration into corporate workflows, not just creative experimentation. The launch discount signals aggressive price competition in the AI video API market, particularly from a major Chinese cloud provider.

Highlights

• 30-second clip generation with claimed improvements in instruction following, inter-shot consistency, and audio quality
• Native support for office documents (DOC, XLS, PPT, PDF, Markdown) as direct creative inputs
• 30% API discount on selected platforms between August 24 and September 23

Critical take

The article reads like a two-paragraph press release summary. It lacks any benchmarks, comparisons with competing models, or independent quality assessments. Consequently, it offers readers nothing beyond Alibaba's own marketing claims regarding consistency and audio quality, leaving a lack of verifiable evidence. The complete absence of technical details (architecture, parameter scale, training methodology) makes it impossible to determine whether Wan3.0 represents a true generational leap or merely an incremental update.