Alibaba released Wan2.7-Image, a unified AI model that handles both image generation and editing within a single architecture. The model introduces fine-grained personalization controls for character features like bone structure and eye shape, precise color specification through color codes, and text rendering support for prompts up to 3,000 tokens. A Pro variant delivers 4K output with a reasoning mode for sharper results.

What Happened

Alibaba Cloud announced Wan2.7-Image on April 1, 2026, positioning it as the next step in the Wan series that previously focused on video generation. Unlike most image models that separate generation from editing into different pipelines, Wan2.7-Image handles both tasks in one unified model. Users can generate images from scratch and then edit specific regions, adjust character features, or modify colors without switching tools.

The personalization capabilities go deeper than typical style transfer. Users can fine-tune specific physical attributes like bone structure, facial proportions, and eye shape to create consistent characters across multiple generations. Color control accepts specific color codes and proportional ratios directly in prompts, giving designers precise palette control.

Why It Matters

Most current AI image workflows require separate models or tools for generation versus editing. Stable Diffusion users, for example, switch between txt2img and img2img pipelines, often losing consistency. A unified model that handles both means fewer handoffs and more coherent results when iterating on images.

The personalization depth matters for creators building characters, mascots, or brand identities. Being able to define and lock specific physical features solves the consistency problem that has made AI-generated characters unreliable for professional use. Combined with color code precision, this targets the gap between creative exploration and production-ready output.

Key Details

  • Unified Architecture: Single model for generation and editing, no pipeline switching.
  • Personalization: Fine-tune bone structure, eye shape, and facial features for consistent character creation.
  • Color Control: Specify exact color codes and color proportions directly in prompts.
  • Text Rendering: Supports prompts up to 3,000 tokens for detailed scene descriptions.
  • Wan2.7-Image-Pro: 4K output resolution with a reasoning mode that interprets and refines prompts before generation.
  • Access: Available now on Alibaba Cloud Model Studio and the Qwen App.

What to Do Next

Try Wan2.7-Image through Alibaba Cloud's Model Studio or the Qwen App. Test the personalization features by creating a character with specific physical attributes, then generating multiple images to check consistency. If you work with strict brand palettes, the color code control is worth evaluating against your current workflow for accuracy.