Alibaba's Wan3.0 generates AI videos up to 30 seconds long from text, images, and documents

Alibaba has launched Wan3.0, the latest version of its video generation model, expanding both the length and input flexibility of AI-generated video clips. Where many competing tools cap output at a few seconds or require a single prompt type, Wan3.0 accepts text, images, PDF documents, and PowerPoint files as source material - broadening the range of practical use cases, particularly for business and presentation-oriented content.

The model supports video generation at up to 1080p resolution and can produce clips as long as 30 seconds, a duration that remains relatively uncommon among publicly available video generation systems. Alibaba has set the price at $6 per 30-second 1080p clip, positioning it as a commercial offering rather than a free research preview. The ability to ingest structured document formats like PDFs and slide decks is a notable differentiator, potentially making it easier for users to convert existing materials into video without manual reformatting.

Wan3.0 builds on Alibaba's ongoing investment in generative AI, a push that has come with significant financial costs. The company reported a 75 percent drop in quarterly profit compared to the same period a year earlier, a decline directly tied to increased AI-related capital expenditure. This pattern - trading near-term earnings for long-term AI positioning - mirrors spending trends seen at other large technology companies investing heavily in model development and infrastructure.

The broader context here is a competitive landscape in AI video generation that has grown considerably more crowded over the past year, with offerings from companies like OpenAI, Google, Runway, and others all targeting different segments of the market. Wan3.0's document-input capability and relatively long clip duration give it a distinct angle, though real-world output quality and consistency will ultimately determine how it fares against established alternatives.