Particle.news

Alibaba Unveils Wan3.0, a 30-Second AI Video Model

Public beta access pairs 30-second, up-to-1080p video with document-to-video conversion at per-second prices below some U.S. rivals.

Overview

  • Alibaba has released Wan3.0 into public beta through Alibaba Cloud’s Model Studio and Qwen Cloud and is preparing a consumer site (wan.video) plus broader API access.
  • The model generates single-pass videos up to 30 seconds long from text, images, audio, video and documents such as PDFs, slides and spreadsheets while keeping characters, objects and layouts consistent.
  • Wan3.0 supports output up to 1080p, creates faces with synchronized micro-expressions, offers multilingual voice output, and converts documents into narrated video sequences.
  • Alibaba lists per-second prices of $0.05 for 480p, $0.10 for 720p and $0.20 for 1080p, putting Wan3.0 below the cited rates for some U.S. competitors.
  • The launch follows a major company funding push for AI and could cut costs for creators and businesses that repurpose documents into video, though real-world performance, safety controls and content provenance will need testing as access expands.