Alibaba's Qwen team has unveiled Qwen-Image-Edit-2511, an upgraded version of its image editing model that brings significant improvements in consistency, character preservation, and built-in support for community-created LoRAs. The model is available on Qwen Chat and via ModelScope for local deployment.
The new version addresses image drift and enhances character consistency, allowing the model to perform imaginative edits on portraits while preserving identity and visual characteristics. Multi-person consistency is also improved, enabling high-fidelity fusion of two separate person images into a coherent group shot.
A key feature is the integration of popular community LoRAs directly into the base model, unlocking effects like lighting enhancement and new viewpoint generation without extra tuning. The model also strengthens geometric reasoning, enabling tasks such as generating auxiliary construction lines for design or annotation.
Practical engineering scenarios are emphasized, including batch industrial product design and material replacement for industrial components. The online version includes speed optimizations, but for best performance, local deployment is recommended.