阿里巴巴开源Qwen-Image模型 可生成吉卜力风格图片

Core Insights - Alibaba's DAMO Academy has open-sourced a new text-to-image model called Qwen-Image, which features 20 billion parameters [1] - Qwen-Image is a MMDiT model capable of generating a wide variety of image styles, including realistic, anime, cyberpunk, sci-fi, minimalist, retro, surreal, and ink wash [1] - The model supports various functionalities such as style transfer, image editing, detail enhancement, text editing, and character pose adjustments [1] - Qwen-Image can also generate images in the popular Ghibli style, showing minimal differences compared to OpenAI's GPT-4o, particularly excelling in understanding complex Chinese prompts and text embedding [1]