Adding the dataset twice as much from v1.1, also means it's 20 times the v1.0 dataset. I also fine-tuned the dataset with a new approach to tag the dataset. The overall look of the generated image is also better than v1.1.