Stability AI unveiled a new version of its image synthesis model on Wednesday. The model, named Stable Diffusion 3 Medium, is designed to convert text requests into images. However, the launch of the model has sparked criticism due to the poor quality of the generated images of people, particularly when compared to similar models like Midjourney and Dall-E 3. It has been revealed that SD3 Medium often produces anatomically incorrect images, leading to ridicule.
On Reddit, a thread titled “This issue is a joke? [SD3-2B]” has emerged, where users are sharing examples of SD3 Medium’s failures in creating images of people, specifically in depicting arms and legs inaccurately.
Additionally, another thread is discussing the challenges faced in generating images of girls lying on grass using the SD3 Medium model.