阿里视频大模型Wan3.0正式上线，行业评价“稳定、真实、有质感”

TL;DR - Alibaba launched Wan3.0, a video-generation model aimed at production workflows, with up to 30-second generation, document-based inputs, and stronger temporal and visual consistency. Its emphasis on stable, realistic output could make AI video more practical for film, advertising, tourism, and music-video production.

• Generates videos up to 30 seconds while maintaining consistency across characters, objects, audio, spatial relationships, and visual style.
• Adds support for DOC, XLS, PPT, PDF, and Markdown inputs alongside reference media.
• Targets realistic rendering of faces, skin, motion, lighting, materials, micro-expressions, and other physical details.
• Available through Alibaba Cloud Model Studio, Qwen platforms, Wan’s website, mobile apps, APIs, and several third-party services.