ShivamShrirao / diffusers

🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch
https://huggingface.co/docs/diffusers
Apache License 2.0
1.89k stars 505 forks source link

generation of multi person in one image is bad. #174

Open tengshaofeng opened 1 year ago

tengshaofeng commented 1 year ago

@ShivamShrirao , Thanks for your great job. I train two person at the same time. one is 'a sks woman', the other is 'a zwx man'. It is ok when using single person, for instance 'a sks woman'. But when I want to create photo of couple, using prompt 'a sks woman and a zwx man', it came bad generation, most time only one person is just so so, the other one is always bad. It seems that when put them together, two person will both be affected. Why? How can I solve that?