Here is Sunny Version 2, trained with a carefully curated 151-image dataset (I discarded a few that felt redundant). This model now represents the Pakura brand character (a favor for a friend), but it’s still fully free to use for any project.

🧾 Base prompts and outfits:

Base prompt:
sunny, 1girl, blonde hair, pink hair, short hair, hair between eyes, eyebrows visible through hair, pink eyes

Lab outfit (standard):
lab coat, collared shirt, black vest, red pleated skirt, black thighhighs, torn legwear,

Lab outfit 2 (more cleavage):
lab coat, red pleated skirt, black thighhighs, torn legwear, cleavage, black vest, torn clothes

Lingerie + lab coat:
red pleated skirt, black belt, black thighhighs, torn legwear, black lace bra, solid black bra, white shirt off shoulders, open shirt, cleavage, bare shoulders, unbuttoned shirt, partially removed clothes, lab coat around arms, seductive smile, casual lingerie outfit

Lingerie + skirt:
red pleated skirt, black thighhighs, torn legwear, black lace bra, cleavage, navel, belt, casual lingerie outfit

Lingerie + stockings:
black thighhighs, torn legwear, black lace bra, black lace panties, cleavage, navel, casual lingerie outfit

Pure lingerie:
black lace bra, black lace panties, cleavage, navel, casual lingerie outfit, barefoot

Bikini (red or black):
(red/black) side-tie string bikini, halterneck, swimsuit

Pakura suit (duck kigurumi):
duck kigurumi, yellow and orange animal costume, duck hoodie, black thighhighs, torn legwear, orange slippers

🧩 Add-ons and features:

• Added visual enhancements like fantasy eyes, moist eyes, glossy eyes, great for close-ups without external LoRAs.

• Fully trained with medium breast tag, making size adjustments more flexible.

• Included new expressions such as fear, disgust, anger, surprise, etc.

• Integrated specific actions like holding sign, ok sign, on back, and more.

• Backgrounds are now more accurate:

  • simple background: pastel solid.

  • white background: pure white.

  • laboratory background: detailed lab setting.

  • beach, ocean, sky, clouds, summer: beach environment.

  • outdoors, ruins, destroyed city, debris, dust, smoke, dramatic lighting, post-apocalyptic, battlefield: apocalyptic ruined city.

Improvements over version 1:

This model addresses previous weaknesses and improves prompt responsiveness.

• Hands, eyes, and back-view coat were refined.

• Outfits are more clearly separated.

• Expressions, poses, and backgrounds that once needed extra LoRAs are now native.

TL;DR:

Version 1 still has its charm and might be better for some tasks, but Version 2 delivers more accurate, consistent, and versatile results with less effort.