Wan 2.7 Reference

• 🎬 Reference-to-video generation from Alibaba's Wan family
• 🎯 Built for strong subject coherence across frames
• 🖼️ Uses reference images to lock character consistency
• 🌍 Part of Alibaba's broad Wan 2.7 media suite
• 🏢 From Alibaba's photorealistic video lineup

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 2.7 Reference is the reference-driven entry in Alibaba's Wan 2.7 video generation suite, released in April 2026 alongside the rest of the 2.7 wave. Where the base Wan 2.7 generates clips from text and Wan 2.7 image-to-video animates a still, this variant accepts one or more reference images to anchor a specific subject — keeping a character, face, or object consistent across frames rather than drifting between shots. That makes it the go-to choice when identity preservation matters as much as motion quality.

The Wan line is Alibaba's photorealistic video family, prized for subject coherence, and the 2.7 generation is the current incumbent — succeeding the earlier 2.6 and 2.5 Preview releases. It sits within a deep Alibaba catalogue that also spans the Qwen text and image models and the newer HappyHorse video family, but for reference-conditioned video Wan 2.7 Reference is the dedicated tool.

It is best suited for creators who need a known subject to stay visually consistent throughout a generated clip — branded characters, recurring personas, or product shots — where a supplied reference image guides the output more reliably than a text prompt alone.