Qwen releases Qwen-Image-Edit-2511 image editor
Alibaba's Qwen team has released Qwen-Image-Edit-2511, an update to its image-editing model that targets image drift and identity consistency, two persistent weak points in generative editing workflows.
The Qwen team announced Qwen-Image-Edit-2511, which it calls an enhanced version of Qwen-Image-Edit-2509. Its headline improvement is better consistency between input and edited images.
According to the team, the main changes are:
- Less image drift and better character consistency, so identity and visual traits are preserved when a portrait is edited.
- Multi-person consistency: 2509 had already improved single-subject edits. The new version extends this to group photos, including merging two separate person images into one coherent group shot.
- Built-in LoRAs: selected popular community-made LoRAs are now part of the base model, so their effects work without extra tuning. The team's examples include realistic lighting control and generating new viewpoints.
- Industrial design: batch product design and material replacement for industrial components.
- Geometric reasoning: for example, drawing auxiliary construction lines for design or annotation.
The model is available in Qwen Chat under the Image Editing feature. The team says the online version is optimized for speed, and it recommends deploying the model locally via ModelScope for the best results.
The announcement has no quantitative benchmarks, no comparisons with other editing models, and no details on model size or licensing. It does not say which community LoRAs were merged. The team's evidence is limited to showcase examples, and it points readers to the existing Qwen-Image Technical Report (arXiv 2508.02324) for citation.