This model was trained on 321 images/1000 step, so in total for 321000 step.

All the images were created with ADetailer and with Hires-Fix.

If you don't want bangs just add it to negative prompts, but keep in mind that it will affect the face quality and she might still have a small bangs on the sides.

THIS IS AN EXPERIMENTAL LORA

Images were gather with a custom script from all sort of social media. Further LoRa will be published with this method to polish training image collecting.

Sources:

https://www.reddit.com/user/puppyavie/

https://linktr.ee/puppyavi