Lates News

This morning, Alibaba DAMO Academy open-sourced the latest Wen-Image model. Wen-Image is a 20 billion parameter MMDiT model that can generate realistic, anime, cyberpunk, sci-fi, minimalist, retro, surreal, ink, and other types of images to support styles transfer, add or delete, enhance details, edit text, and adjust character poses, among other common operations. Wen-Image can also generate Studio Ghibli-style images, which have become popular all over the Internet thanks to OpenAI's GPT-4. According to tests, the difference between the two models is very small, especially in understanding highly complex Chinese prompts and text embedding, where Wen-Image performs better (AIGC open community).