Lora trained on images of Ana de Armas, Natalie Portman, Elizabeth Olsen and Emma Watson. With 254 images for Ana de Armas, 199 images for Elizabeth Olsen, 175 images of Emma Watson and 137 images for Natalie Portman. All images are first captioned by GPT-4 Vision and then aligned by me to allow trigger wrds to have their desired effect. The Lora strength should by around 0.8-1.2. But you should experiment on your own to see what you like the best. The model is definitely heavily biased towards realism, but will give appropriate results with a checkpoint that has been tuned towards Anime etc. The trigger words are just the names of the women: Elizabeth Olsen, Ana de Armas, Emma Watson and Natalie Portman.
