Alibaba launches Wan2.7-Image to open a new era of individualized appearance

Ali's large model team officially released the unified image generation and editing model Wan2.7-Image today. Compared to its predecessor, this model has achieved a qualitative leap in portrait customization, color control, and long-text rendering, aiming to break the common "one-size-fits-all" feeling in AI-generated images.

Currently, the model is available on the Alibaba Cloud BaiLian platform via API, and it is also open for experience on the WanXiang website.

Core Upgrades: Virtual "Facial Sculpting" and Precise Color Palette

Wan2.7-Image introduces several industry-leading differentiated features, greatly enhancing creative freedom:

• Unique Faces for Each Person: It enhances the virtual character "facial sculpting" capability. Users can precisely control bone structure, eye shape (such as almond eyes or phoenix eyes), and facial details through prompts, completely eliminating standardized AI faces.

• Precise Color Palette: It newly supports color control functionality. Users can extract the color ratio from reference images, accurately replicating colors like Van Gogh's bright yellow or Picasso's cool blue into new works.

• 3K Token Ultra-Long Text Rendering: It solves the problem of AI text generation, supporting up to 12 languages, and can render complex text, tables, or formulas the size of an A4 page with print-quality results.

In the field of image editing, Wan2.7-Image introduced an "interactive editing" feature. Users can simply select a precise area to add, align, move, or even perform pixel-level logical replacement (e.g., replacing ice cubes with fruits while keeping the environment's lighting unchanged).

Additionally, the model supports generating groups of up to 12 images, ensuring high consistency in style and characteristics among multiple subjects (such as group photos or furniture combinations) across different scenes.

Ali stated that Wan2.7-Image adopts a unified model architecture for generation and understanding. By achieving semantic mapping in a shared latent space, the model no longer blindly guesses pixels corresponding to text but truly possesses underlying semantic understanding.

In the 2026 visual creation competition, the release of Wan2.7-Image indicates that AI painting is evolving from "random card-drawing" generation to "industrial-level" precision control. Whether for short film storyboards, e-commerce advertisements, or social transformation, this high-precision editing capability will significantly lower the threshold for professional content production.