Wan 2.5 Preview

• 🎬 Alibaba's text-to-video generator for photorealistic clips
• 🏢 Part of Alibaba's Wan video family
• 🎯 Strong subject coherence across frames
• 🎨 Built for cinematic, lifelike motion
• 💬 Generates video directly from text prompts

Alibaba

Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research,…

Wan 2.5 Preview is Alibaba's text-to-video model in the Wan family, generating photorealistic clips from text prompts with an emphasis on subject coherence — keeping characters and objects stable and consistent as frames progress. Released in September 2025, it arrived as an early preview alongside the companion Wan 2.5 Preview image-to-video variant and the Wan 2.2 A14B release, giving creators a text-driven path into the Wan generation pipeline.

Within Alibaba's lineup it represents an earlier generation of the text-to-video line, since superseded by Wan 2.6 (December 2025) and the current incumbent Wan 2.7 (April 2026). The broader Wan catalogue now spans image-to-video, reference-to-video, editing, and Pro tiers, reflecting how the family has expanded well beyond this preview's scope.

As a preview build, Wan 2.5 is best suited to experimentation and short photorealistic sequences where frame-to-frame consistency matters more than the latest features. Users seeking newer capabilities or production polish should look to the later Wan releases, but this model remains a straightforward option for text-prompted video generation.