Kyarine Cath PonyXL (Final Fantasy XIV OC) | LoCon

Kyarine Cath my Warrior of Light from the MMO "Final Fantasy XIV". Works best around 0.8-1.0 weight. Training done on Pony Diffusion V6 XL. Request form

V2:
Always needed: "Kyarine, dragon girl, grey skin, horns, ringed eyes, freckles, scales, short hair, streaked hair, two-tone hair, black hair, spiked tail"

• Add "medium breasts, grey nails, toned" for accuracy.

  • Note "muscular" can look better on some checkpoints but is overkill on others.

Suggestions/notes:

• Remove "dragon tail" if you want a more upper body/portrait shot.

• For outfits go have fun as it was trained to be fully flexible.

  • I suggest lower the weight a little if the scales are bleeding through outfits.

• Flexible and very NSFW able.

• Goal of V2 was to reduce the 3d style burn and add quality/stability with a new dataset from Dawntrail with the graphics update.

  • Trained on 54 new images.

V1:

Trigger: "Kyarine, colored skin, grey skin, grey lips, orange eyes, freckles, dragon girl, dragon horns, black scales, short hair, spiked hair, streaked hair, black hair, dragon tail, toned" always needed for accuracy.

• "medium breasts" suggested for accuracy.

• Add "abs, navel" where visible for accuracy.

• I highly recommend "3d" in the negative to avoid the FF14 3d style.

Suggestions/notes:

• Remove "dragon tail" if you want a more upper body/portrait shot.

• If hair streaks fail add "orange hair" but seems to work nicer without this tag.

• For outfits go have fun as it was trained to be fully flexible.

  • I suggest lower the weight if the scales are bleeding through outfits.

• Flexible and very NSFW able.

My warrior of light from back when I started final fantasy 14 about 5 years ago and I never changed her outside of hairstyles. Plus I like fit ladies so training her was biggest yes and now that Pony is so good I was able to get her working real nice considering I only used 3d in-game screenshots.

I hope you all enjoy, have fun with her :)

Feedback and reviews are always appreciated.

Training data is in the Lora metadata.

Trained on 60 images.