Wan 3.0 Prime Pro Reference

• 🏢 Alibaba's Wan 3.0 video family, reference-to-video variant, released September 2026.
• 🖼️ Generation guided by reference material rather than a prompt alone.
• 📏 Wan3.0 generates up to 30 seconds in a single pass.
• 👁️ Family supports 480p, 720p and 1080p output resolutions.
• 🔧 Wan3.0 unifies text, image and reference conditioning in one family.
• 💬 Documented API inputs include prompt, resolution, duration and seed.
• 🆕 Prime Pro is the Pro variant within the Wan 3.0 Prime line.
• 🌐 Served through Alibaba Cloud Model Studio's Wan3.0 video endpoints.

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 3.0 Prime Pro Reference is the reference-to-video entry point into Alibaba's Wan 3.0 generation, sitting alongside Wan 3.0 Prime Pro (text-to-video) and Wan 3.0 Prime Pro (image-to-video) for text and image conditioning. Instead of working from a prompt alone, it takes supplied reference material and uses it to condition the generated clip.

Alibaba positions Wan3.0 as the latest generation of its video family, unifying text-to-video, image-to-video (first frame or first-and-last frame) and reference-based generation, with 480p, 720p and 1080p output and a maximum duration of 30 seconds in one pass. That documented single-pass length is one of the defining structural traits of this generation relative to earlier Wan releases such as Wan 2.7 Reference and Wan 2.6.

Within Wan 3.0, Prime is a distinct documented model line in Alibaba Cloud Model Studio, and Prime Pro is the Pro variant of that reference-to-video line, offered in the catalog alongside Wan 3.0 Prime Reference and Wan 3.0 Reference.

On the API side, Model Studio's Wan3.0 video generation reference documents an asynchronous task-based workflow, with parameters covering the prompt, output resolution, clip duration and seed. Catalog metadata describes the family as aimed at photorealistic generation with strong subject coherence across frames.