Trained entirely on MLP art consisting of 270 images of backsides and 79 images of portraits where the character is looking back. Mostly close-ups with a good mix of body shots, approximately 10% male focus. Trained with Booru tags.
The Illustrious version is trained on IllustriousXL and is the most flexible, capable of working with human, anthro, and pony characters. Not as good at recreating existing mlp characters, but a little better at furry than stock Illustrious. This version is the best for human characters, very good at putting furry tails on humans.
The Pony version is trained on PonyXL and offers significant enhancements over the SD 1.5 version. It can render characters very faithfully, supports a wider range of poses, and it's generally much better at generating male characters. Not limited to existing pony or feral characters, but non-pony results may be less successful. Improves the quality of PonyV6 by basically all jpg artifacts.
The SD 1.5 version is ok, but it does struggle to recreate existing mlp characters, and it's the least flexible version. It works best when prompting heavily for an ", looking back" style prompt.
Helpful Positives: my little pony, puckered , , , , looking back, colored skin, tail, long hair, very long hair, wings, unicorn, single horn, kneepits,
Helpful Negatives: gaping, spreading
Retrained using Illustrious as base.
A future update may include a focus on human characters.
No dataset changes.
Images made with LORA weight 1.0
Trigger word: my little pony, puckered
1. 轉载模型僅供學習與交流分享，其版權及最終解释權归原作者。
2. 模型原作者如需認領模型，請通過官方渠道联系海藝AI工作人員進行認證。我們致力於保護每一位創作者的權益。 點擊去認領
