News on August 19,Thousand Questions on Tongyi Qwen Today announced the launch of Qwen-Image-Edit -- Qwen-ImageImage EditingVersion.
Qwen-Image-Edit is based on the Qwen-Image model of 20B, which has been further trained to extend the unique text rendering capability of Qwen-Image to image editing.Enables precise editing of text in images.
In addition, Qwen-Image-Edit inputs images to both Qwen 2.5-VL (for visual semantic control) and VAE Encoder (for visual appearance control).Thus, it has both semantic and appearance editing capabilitiesUsers can visit Qwen Chat (chat.qwen.ai) and select the "Image Editing" feature. Users can visit Qwen Chat (chat.qwen.ai) and select the "Image Editing" feature to experience it.
Key features of Qwen-Image-Edit include:
- Dual editing of semantics and appearance:Qwen-Image-Edit supports not only low-level visual appearance editing (e.g., adding, deleting, modifying elements, etc., which requires the rest of the image to be completely unchanged), but also high-level visual semantics editing (e.g., IP authoring, object rotations, style migrations, etc., which allows for pixel-by-pixel variations but maintains semantics consistency).
- Precision text editing:Qwen-Image-Edit supports Chinese and English bilingual text editing, which can directly add, delete, change and other operations on the text in the image while keeping the original font, font size and style.
- Powerful benchmark performance:Evaluations in several public benchmarks show that Qwen-Image-Edit delivers SOTA performance on image editing tasks and is a powerful base model for image editing.
One of the highlights of Qwen-Image-Edit is its dual editing capabilities of semantics and appearance. The so-called semantic editing refers to the modification of image content while keeping the visual semantics of the original image unchanged.
Application Scenarios: From Creative Design to Commercial Realization
The versatility of Qwen-Image-Edit makes it suitable for a variety of scenarios, including but not limited to.
- Poster and advertising design:: Generate visually appealing promotional posters with support for complex text typesetting and style migration.
- IP Content Creation: Generate MBTI-themed emojis based on brand mascots (e.g. Qwen's Capybara) to maintain character consistency.
- Education and training:: Quickly generate high-quality illustrations and charts to enhance the visual appeal of course content.
- Games & Movies:: Support for character design, background generation and new perspective compositing to optimize the asset development process.
User feedback shows that Qwen-Image-Edit's intuitive operation and high quality output make it an ideal tool for non-professional designers. For example, one content creator said, "Qwen-Image-Edit allows me to complete my marketing visuals in minutes, with accurate text rendering and results comparable to professional software."
1AI Attached open source address:
- ModelScope: https://modelscope.cn/models/Qwen/Qwen-Image-Edit
- Hugging Face: https://huggingface.co/Qwen/Qwen-Image-Edit
- GitHub: https://github.com/QwenLM/Qwen-Image