V1 – First public release
Training: 64 images (SFW + NSFW), auto-label with fixed prepend tags, 15 epochs. Chosen Epoch 9 for best balance of consistency and flexibility.
Target style: Cartoon / cel-shaded, clean outlines, two-tone afro (brown + blonde streaks), green eyes, freckles.
Compatibility: SDXL (recommended base: WAI-NSFW-Illustrious SDXL or similar anime/cartoon checkpoints).
Recommended weight:
Prompt snippets (examples):
SFW: solo, full body, standing, hands on hips, confident smile
NSFW: nude, kneeling, arms behind back, blushing
(Combine with your fixed character block.)
Negatives (recommended):
multiple characters, partner, male, duplicate, watermark, text, leaf, sunglasses, extra accessories
Known quirks:
Occasional leaf/accessory appears → suppress via negative tags (see above).
Slight front-pose bias. For variety, clearly state pose (sitting sideways, looking over shoulder, one leg raised, etc.).
For a shorter/petite body: add petite, short girl, small frame.
Roadmap v1.1:
- Short retrain (3–5 epochs) with filtered captions (no accessories), more pose diversity, additional solo NSFW poses.
Enjoy – feedback and example generations will help fine-tune the next version!
