CURRENTLY RETRAINING TO FIX A BUG. The model is still usable but you may run into a random issue where the image is "squished". I believe its because my dataset had images that were too wide. Currently retraining with a manually cropped dataset. Will update to 1.1 when I wake up.

Just prompt Kamala Harris. For best results use the full Flux1.D model and fp16 clip. FP8 model works as well. Only trained 1,600 steps because the base model at least had an idea of what she was. The likeness is increased for sure though. Can always bump up the training more if needed. Post results and let me know what you think. Will upload better sample images when I return home from work.