V4 is the same training data set as V2, however the first two times I trained the model I used images that had be rescaled with a "film grain" filter added. This created a style within the LoRA that made even the background have a graininess texture too it. I resized the original images without the filter and retrained. Smooth texture unless prompted now.

This LoRA is an experiment in translating a 2D animated style into photorealism. Using the structural data of Krea2 animated characters knowledge, I utilized targeted prompt engineering to generate photorealistic variations of the character while maintaining facial stability and feature consistency.

From that generation run, I curated a dataset of roughly 40 images to train this model on Peggy Hill from King of the Hill (1997) who is canonically in her early 40s. This specific version is trained at 1,000 steps, though testing was conducted up to 2,000 steps. Because the source material is a highly stylized 2D cartoon, the measure of success and interpretation of a "realistic" counterpart is inherently subjective. I welcome your feedback and test images!

UPDATE v2: I generated a new image set focused on lowering the approximate age to 35-40 years old. The hair style can still occasionally age the character, but it is significantly better. When generating the new training images, I used text prompting to add a taut post-partum belly with stretch marks. These "stretch marks" do render when the stomach is exposed. The character's bouffant hair-style is much more consistent and a better match the original Peggy Hill character. Finally the term "young woman" can be used to age the character to 25-30 years old (see the pregnant example image to view it's effectiveness). This version required 2000 steps to hit the mark.

UPDATE: I'm working a new set of images to use for training to move the character closer to late 30s or early 40s. I've found the best way of changing this is through prompting by using 20-something, 30-something, or 40-something (or older if that's your preference). Those tags will adjust the appearance of the character and will try to post some images of the affects.