Wan 2.7 Reference

• 🎬 Reference-guided video generation from Alibaba's Wan family
• 🏢 Built by Alibaba, the Wan 2.7 generation
• 🎯 Strong subject coherence across frames
• 🎨 Photorealistic output with consistent characters
• 🖼️ Uses reference images to steer generation

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 Alibaba's reference-to-video model, generating photorealistic clips that preserve a subject's appearance across frames by conditioning on supplied reference imagery. Released in April 2026, it sits within the broader Wan 2.7 generation alongside the standard text-to-video and image-to-video entries (both named Wan 2.7), the Wan 2.7 Edit video transformer, and the Wan 2.7 Pro image generator — a full multimodal media suite from Alibaba's cloud research efforts.

Within the Wan lineup, this is the dedicated reference-driven member, distinct from the prompt-only and single-image workflows. Its emphasis on subject coherence makes it the variant to reach for when a specific person, character, or product must stay visually consistent throughout a generated sequence, rather than drifting between shots.

In practice, Wan 2.7 Reference is best suited to creators who need controllable, identity-stable video — character animation, branded content, or any scenario where matching an existing visual reference matters more than open-ended generation. Pair it with the other Wan 2.7 tools for editing or still-image work as a project demands.