Alibaba introduces Qwen Image Edit to perform AI-powered image edits in seconds

Alibaba has introduced Qwen Image Edit, expanding the capabilities of its Qwen Image model into image editing applications. Built upon the 20 billion parameter Qwen Image, this new model brings precise text rendering technologies to image editing, enabling accurate text adjustments within visuals.

To facilitate comprehensive editing, Qwen Image Edit processes input images through both the Qwen2.5-VL model, which manages visual semantics, and a VAE Encoder, which controls visual appearance. As a result, users can perform low-level visual appearance edits — such as adding, removing, or modifying elements in an image without affecting unchanged regions — alongside high-level semantic edits, like object rotation, intellectual property creation, and style transfer that maintain semantic consistency despite pixel changes.

Following these technical advances, Qwen Image Edit also supports direct addition, deletion, and modification of bilingual text (Chinese and English) in images. It preserves the original font, size, and style, making it relevant for multilingual projects. Benchmark evaluations indicate Qwen Image Edit delivers state-of-the-art performance compared to existing image editing solutions. Qwen Image Edit is anticipated to help lower technical barriers and foster further innovation in visual content creation.