Alibaba Releases Wan3.0 in Public Beta: 30-Second AI Videos from Text, Images, Audio, or Documents

My take: Alibaba just raised the ceiling on what you can generate in a single AI video pass. Wan3.0 produces clips of up to 30 seconds in one shot, twice the 15-second limit of its previous version. What makes it different isn't only the length: it accepts text, images, audio, video, PDFs, PowerPoint presentations, spreadsheets, and web pages as starting points. You can hand it a corporate slide deck and ask it to turn it into video.

For content creators, training teams, and anyone producing explanatory material, this simplifies a process that previously required multiple tools. The price at 1080p is $0.20 per second, about $6 for a full 30-second clip. It's available in public beta on Alibaba Cloud Model Studio and Qwen Cloud since August 6, though API access is still by invitation.

Model weights are not publicly available yet, so you can't run it locally for now. It's worth watching closely: if open access arrives, the use case for content production at scale is immediate.

What types of documents or presentations in your work could gain real value if turned into video automatically?