Qwen Image Edit TL5
Qwen-Image-Edit is Alibaba's open-weight image-to-image editing model, built on the 20B-parameter Qwen-Image diffusion transformer. It takes an input image and a text instruction and produces a revised image, combining Qwen2.5-VL for semantic understanding with a VAE encoder for appearance control. The model supports localized edits such as adding, removing, or changing objects while preserving unaffected regions, as well as broader transformations including style transfer, object rotation, and character variations. A standout capability is editing text already embedded in images: it can add, replace, or delete English and Chinese words while aiming to retain their font, size, and styling. Applications include revising posters and ads, updating signage, adapting illustrations, and iterating on product visuals. Released under Apache 2.0.
Qwen Image Edit TL5
Qwen-Image-Edit is Alibaba's open-weight image-to-image editing model, built on the 20B-parameter Qwen-Image diffusion transformer. It takes an input image and a text instruction and produces a revised image, combining Qwen2.5-VL for semantic understanding with a VAE encoder for appearance control. The model supports localized edits such as adding, removing, or changing objects while preserving unaffected regions, as well as broader transformations including style transfer, object rotation, and character variations. A standout capability is editing text already embedded in images: it can add, replace, or delete English and Chinese words while aiming to retain their font, size, and styling. Applications include revising posters and ads, updating signage, adapting illustrations, and iterating on product visuals. Released under Apache 2.0.