In the version currently being tested, I think the facial features look more animal-like than in the beta version, which is a significant improvement.
V1(a) uses regularized images (twice the volume of the original training images), while V1(b) does not use regularized images.
Include these prompts at a minimum :
furry head, furry face, goat head man, furry man, white horns, long horns, goatee, pointy ears, glowing eyes, yellow eyes, no pupils, fur skin, dark blue gray skin, dark blue gray face, wavy hair, semilong hair, wavy bangs, messy bangs, wavy sidelocks, side locks on jaw, green hair, goatee, masculine face, colored skin,
Clothes:
collared cape, high collar, purple cape, purple loincloth, yellow rope with chests, simple cape, loose loincloth, simple loincloth,
(Beta only)Clothes(Playing manual):
clothing, (gray robes:1.4), long sleeve, purple cape, purple loincloth, loose loincloth , black pants, loose pants, green shoes, green copper shoes, reg slit robes, gold bracelet:1.5).
(I'm forcing the manual version's clothes onto the packaged version of him, so this one doesn't look right either.)
Description
Testing now...
This model uses twice as many regularized images created in the beta version as it does original training images.