Generate variation in images using GAN based network

I need a GAN-based model which can generate different images based on seasons(winter, summer, rainy, windy, etc) on the BDD100K dataset. I need the model to help me generate a new dataset with the existing image dataset such that they look like real captured images. There is a Github repo that can be helpful (https://github.com/EndlessSora/TSIT). We can use the repo and make changes to it so that it works only specifically for BDD100K image dataset. The repo can do image variation on sunny images but I want to work for rainy, snowy, etc images as well.