Alibaba’s next iteration of Wan-based image AI model is here, and it is the Wan2.7-Image, touted to fix what’s been holding AI image generation back for a while now.
With current AI models often generating generic outputs with weird colors, inconsistent details, and even the “always the same face and portrait angle”, Wan2.7-Image offers even more through fine-tuning in things like bone structure or eye shape to create more realistic and unique characters. On top of that, there’s a built-in color palette system where you can input exact color codes and proportions, which makes it way easier to match brand guidelines or replicate specific artistic styles without guesswork.
Text rendering has also been improved, as Wan2.7-Image can now support up to 3K tokens and 12 languages, leading to clearer text for academics, formulas, and even tables. Its batch processing capability can take in 9 reference images and output 12 at once, making things like storyboards, architecture concepts, or even full e-commerce campaigns easier than ever.
They also added a “click-to-edit” system, which basically lets you select parts of an image and adjust them directly – moving elements around or refining details with pixel-level precision.
Behind the scenes, the model is trained to better understand context – not just pixels, but things like lighting, composition, and layout – so it can interpret prompts more accurately instead of just approximating them.
And if you want something more advanced, there’s also Wan2.7-Image-Pro. That version brings more stable compositions, sharper prompt understanding, and full 4K output, which clearly targets more professional use cases.
As for availability, Alibaba is rolling this out through its Model Studio platform, its official Wan website, and it’s also being integrated into the Qwen App.
