Eve — three outfits, one trigger, trained for MOSS
The Diving Suit, the Skin Suit and Racer’s High, each callable by name, all in one LoRA.
This is a character LoRA for MOSS Image. MOSS honestly picked up the training data better than any model I’ve worked with: all three outfits hold, including Racer’s High, which had only a handful of reference images.
How to use
Start every prompt with the identity line:
OnlyEve, 1girl, solo, black hair, very long hair, high ponytail, blunt bangs, hair ornament, brown eyes, large breasts
Then add one outfit:🟢 Diving Suit (the default)
eve diving suit, green and white bodysuit, skin tight, sleeveless, high collar, green necktie, white fingerless gloves, green armbands🤍 Skin Suit
eve skin suit, beige skin tight bodysuit, latex, white geometric panels, laced spine down her back, highleg, wrist cuffsTip: add from behind, looking back. The laced spine is the best part.
🏁 Racer’s High
eve racers high outfit, red and black checkered cropped jacket, long sleeves, black latex bodysuit with cutouts, criss-cross red strings, hanging chains, black gloves hip vents sideless,Tips
Strength: start at 1.0. Lower it if the image looks too rendered.
Mixing outfits: the outfit names are separate, so you can also mix the parts (a checkered jacket over the diving suit, anyone?).
Style: the dataset is fan art from many artists plus a few 3D renders. Each image’s style is captioned, so the LoRA leaves the art style to your model or style LoRA. Pair it with whatever look you like.
Prompting: MOSS reads sentences. Use tags for her and her outfit, and plain sentences for the pose, the place and the light.
Not great at
Racer’s High from unusual angles. It had the fewest training images.
Front views of the Skin Suit are a bit less exact than the back.
Other outfits from the game aren’t trained, so asking for them gives you a guess.

.jpeg)