WAN 2.7
Think First, Then Generate
WAN 2.7 by Alibaba brings reasoning to image generation. It reads your prompt, plans the composition, then renders — so complex instructions land the way you meant them. Text-to-image and image editing live in one model, available in Standard and Pro tiers.
Capabilities
Standard & Pro tiers, reference-based editing
Key Features
Thinking Mode reasoning, multi-reference consistency, unified generation and editing
Thinking Mode — Plans Before It Paints
WAN 2.7 doesn't jump straight to pixels. It interprets your prompt, reasons through composition and layout, then generates the final image. The payoff is fewer misread instructions and tighter adherence to complex, multi-part briefs.
Multi-Reference Consistency
Feed WAN 2.7 up to nine reference images and it holds the same subject, character, or style across every output. Lock a face, a product, or an art direction once, then generate variation after variation that still belongs to the same shoot.
Generation + Editing in One Model
One engine creates images from a blank prompt and reshapes existing ones — no switching tools between steps. Restyle a photo, composite elements from several shots into one scene, or refine a result, all without quality loss between generation and editing.
Compare Models
| WAN 2.7 | WAN 2.7 Pro | Qwen Image 2.0 | |
|---|---|---|---|
| Developer | Alibaba | Alibaba | Alibaba |
| Architecture | Unified Gen + Edit, Thinking Mode | Unified Gen + Edit, Thinking Mode | Unified Gen + Edit (7B) |
| Max Resolution | Up to 2K | Up to 4K | Up to 2K |
| Speed | Fast | Quality-first | Fast (7B) |
| Image Quality | |||
| Photorealism | |||
| Text Rendering | |||
| Editing | Gen + edit, compositing (up to 9 refs) | Gen + edit, compositing (up to 9 refs) | Full editing (up to 3 refs) |
| Best For | Complex prompts, multi-reference | Highest-fidelity 4K output | Professional, text-heavy |
Start creating with WAN 2.7
WAN 2.7 is available with any Workroom subscription. Pick a plan and start generating.
