Use <code>`Pottsness`</code> in your prompt for best results. Use 0,8-1,2 strength, with 1,0 being the sweet spot for reducing stubbornness and keeping overall likeness to the artist. While 1,2 is the most like to the artists overall, and gives higher quality results. Use 3.5 distilled CFG. (3,5=1 CFG) Begging Please, please like an image of mine or post your results! lol. Training Details This is very Low Rank LoRA. It was trained on a dataset of 5 images from Pottsness with 20 repeats and 10 epochs total for training across the dataset. I forgot to include <code>artist's name</code> on certain images. This has altered the flux font and caused writing and text bubbles to appear. Sorry. (._. ); If anyone requests, I will retrain it. This is my tenth (an best) attempt at this LoRA using different dataset variations. Half of this failure rate is due to poor cloud computing, and not data or settings.

Pottsness Style | FluxD | [Experiemental]

Has a few issues. Could be better, even with its limited training data.

Files with SHA256: df4604b34a6421c06ec5c2e27a74970c9f6700adfef3eea2d22693bae2418649