v2: I used google ai to prompt the images where it also describe the lips type like full lips, smooth lips, big lips and normal lips. this lora is less focus on dalle3 style and more on general styles.

this lora main focus is on lips, like pressing lips, many base model and tagging model don't have any concept for lips or is not good with lips in general, So I needed to manually tag every images.

I tried to train the model to make a biting lips image, but it failed. Every sdxl model I tried is bad at it. You can try to see if you can make a image where the character is biting his lips.

if you are like me, you probably going to like this lora.

I took small amount of images from my dalle3 dataset (around 1000 images) and took 150 images that focus on lips.