lorelai blyndeff

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:

• All-round: 0.8–1.0

• Max character accuracy: 0.95–1.1

• Style mixing: 0.6–0.7

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!