Alibaba Group said on Wednesday it has launched Wan 2.7 Image, an AI model specialised in image generation and editing.
Wan 2.7 Image is designed to address uniform styles and unstable colour output, which have been cited as limitations of existing AI image generation models. Users can adjust details such as skeletal structure and eye shape to create differentiated characters by project. Through a colour palette function, users can precisely reflect brand colours by entering specific colour codes and ratios in prompts.
Text rendering has also been improved. It supports text inputs of up to 3,000 tokens and can generate academic text and complex formulas and tables in print quality in 12 languages. It also includes batch processing that can use up to 9 reference images and generate up to 12 images at a time.
A "click editing" interface also allows users to select specific areas of an image and add, move or align elements at the pixel level.
Alibaba also unveiled a pro model, which it said supports stable image composition, improved prompt understanding and high-resolution 4K output. The two models are also set to be integrated into the Qwen AI application app.
Seulgi Son
sageson@d-today.co.kr