I know most of those generative image models do a bad job with English words/Latin letters; I wonder if it’s just as bad with Chinese characters?