Alibaba open-sources the Qwen-Image model to generate Ghibli-style images
2025-08-04 22:50:55
On August 5th, early this morning, Alibaba Dharma Academy open-sourced the latest Text-to-Image model Qwen-Image. Qwen-Image is an MMDiT model with 20 billion parameters, which can generate dozens of types of pictures such as realism, animation, cyberpunk, science fiction, minimalism, retro, surreality, ink painting, etc. It supports style transfer, addition and deletion, detail enhancement, text editing, character posture adjustment and other routine operations. Qwen-Image can also generate Ghibli-style pictures of OpenAI's GPT-4o explosion network. It is tested that the gap between the two is very small, especially in the understanding of ultra-complex Chinese prompt words and text embedding. Qwen-Image is better.
Disclaimer:
1. The information provided does not constitute investment advice. Investors should make independent decisions and bear all risks themselves.
2. The copyright of this content belongs to the original author. The views expressed herein are solely those of the author and do not represent the stance or position of this website.
