Qwen Image Edit - SOTA Open Weight Image Editing Model
6 days ago
- #image-editing
- #AI-models
- #Qwen-Image
- Qwen-Image-Edit is an image editing model built upon the 20B Qwen-Image model, extending its text rendering capabilities to image editing tasks.
- The model supports both semantic editing (IP creation, object rotation, style transfer) and appearance editing (adding, removing, modifying elements while keeping other regions unchanged).
- Qwen-Image-Edit excels in precise bilingual (Chinese and English) text editing, preserving original font, size, and style.
- The model achieves state-of-the-art (SOTA) performance in image editing benchmarks.
- Quick start guide includes installation of diffusers and a code snippet for generating images based on text prompts.
- Showcase examples include semantic editing (Capybara mascot transformations, MBTI-themed emoji packs, novel view synthesis, style transfer) and appearance editing (adding signboards, removing fine hair strands, modifying text color).
- The model's accurate text editing capability is highlighted with examples of editing both English and Chinese text, including calligraphy corrections.
- License agreement under Apache 2.0 and citation details are provided.