Alibaba Group's Qwen team has made another breakthrough in the field of AI vision, officially launching the upgraded image editing model Qwen-Image-Edit-2511. To address the common issues of "facial distortion" or "loss of identity" when using AI for photo editing, this model represents a significant leap forward, ensuring accurate preservation of facial features in the original image while allowing creative modifications.

image.png

According to the official information released by Qwen on Hugging Face, this model is a major update from its previous version. It not only handles fine-grained editing of single-person portraits but also performs well in complex scenarios such as group photos, maintaining the consistency of identities among multiple characters. Whether it is changing clothing, adjusting the background, or modifying lighting effects in a photo, the model can achieve natural transitions without altering the person's "appearance."

In addition to the core identity preservation technology, the new version of the model has enhanced control over camera perspective, geometric calculations, and industrial product design. Notably, the Qwen team has directly integrated the widely popular LoRA (fine-tuning model) technology into the base model, making it easy for ordinary users to achieve professional-level editing results.

huggingface:https://huggingface.co/spaces/Qwen/Qwen-Image-Edit-2511

Key points:

  • 👤 Perfect Identity Retention: The new model solves the common "face-changing" issue in AI photo editing, perfectly maintaining identity consistency during creative edits.

  • 🛠️ All-Round Editing Tool: Not only good at portrait processing, it also enhances control over lighting, industrial design, and multi-object collaborative editing.

  • 🔓 Open Source and Free to Use: The model is open-sourced under the Apache 2.0 license and provides a web-based Demo for users to test directly.