<h3 id="heading-1243">weight 0.7 is best.</h3><h3 id="heading-1244">☆Very important swap <code>blue eyes</code> ↔ <code>black blindfold, covered eyes</code> <code>black dress, black skirt </code>↔ <code>covered navel, (white leotard:1.2)</code></h3>outfit A (blindfold, dress) <code>hm2b, black blindfold, covered eyes, mole under mouth, clothing cutout, long sleeves, puffy sleeves, juliet sleeves, feather trim, black thighhighs, black gloves, black dress, black skirt</code>outfit B (blindfold, leotard) <code>hm2b, black blindfold, covered eyes, mole under mouth, clothing cutout, long sleeves, puffy sleeves, juliet sleeves, feather trim, black thighhighs, black gloves, covered navel, (white leotard:1.2)</code>outfit C (without blindfold, dress) <code>hm2b, blue eyes, mole under mouth, clothing cutout, long sleeves, puffy sleeves, juliet sleeves, feather trim, black thighhighs, black gloves, black dress, black skirt</code>outfit D (without blindfold, leotard) <code>hm2b, blue eyes, mole under mouth, clothing cutout, long sleeves, puffy sleeves, juliet sleeves, feather trim, black thighhighs, black gloves, covered navel, (white_leotard:1.2)</code><pre><code>lora Ver.2

144MB → 36MB
Remove low-quality datasets</code></pre><h3 id="heading-3"> If you use this LoRA in a reality model (Checkpoint), it may not be perfect.</h3>

Files with SHA256: a14135f959383d53feb825c5b57734c0a2865eff339f5aff7df39c258ba2f1fd