As per the official GitHub post on the AI tool, the new edit feature builds on Qwen-Image, a 20-billion-parameter foundation model released earlier this month. Screenshot: Desk
“>
As per the official GitHub post on the AI tool, the new edit feature builds on Qwen-Image, a 20-billion-parameter foundation model released earlier this month. Screenshot: Desk
Alibaba’s Qwen research group has released an open-source AI model, Qwen-Image-Edit, which enables advanced image editing through text prompts.
As per the official GitHub post on the AI tool, the new edit feature builds on Qwen-Image, a 20-billion-parameter foundation model released earlier this month.
For all latest news, follow The Daily Star’s Google News channel. According to the post, the model can handle both semantic edits – broad changes that alter style or structure without losing the identity of an image – and appearance edits, which focus on precise, local modifications.
A key feature of Qwen-Image-Edit, according to Alibaba, is its ability to work with text inside images in both English and Chinese. It can add, remove, or correct writing while preserving font, size, and layout. Demonstrations in the official source have shown its use in correcting classical Chinese calligraphy, refining signage, and generating themed emoji packs based on Qwen’s capybara mascot.
The system also supports functions like style transfer, object rotation to reveal unseen angles, background changes, and content insertion or removal. These applications range from advertising and design to cultural preservation and personal photo editing.
Qwen-Image-Edit is available under an Apache 2.0 licence and can be downloaded for local use, deployed on cloud services, or accessed through platforms such as Hugging Face, ModelScope, GitHub, and Qwen Chat. Enterprises can also use the tool through Alibaba Cloud’s Model Studio.
