Itsuki Takao from Hazurewaku no "Joutai Ijou Skill" de Saikyou ni Natta Ore ga Subete wo Juurin suru made (ハズレ枠の【状態異常スキル】で最強になった俺がすべてを蹂躙するまで) This model was trained on anime screencaps and several official arts.<pre><code>Note:
This is a B-LoRA. If you want to replicate the character in anime style, use both LoRAs together and then add anime screencap as a positive prompt. You may also need to avoid using all of the score tags.</code></pre> Prompts:-<ul><li>Trigger Words:</li></ul><pre><code>itsuki_olis, 1girl

Note: This is my first time removing hair and eyes tags from my dataset, meaning that you don't need to include them in the prompts. However, I realized that this could reduce the accuracy of her hairstyle, hair color and eye color. Using the style LoRA with a medium weight along with the character LoRA can help improve this.</code></pre><ul><li>Main Outfit:</li></ul><pre><code>black scarf, cleavage, cropped vest, midriff, groin, highleg, white shorts, elbow gloves, arm guards

Extra: (not very accurate)
black thighhighs</code></pre><ul><li>School Uniform: (not very accurate)</li></ul><pre><code>white shirt, black skirt, loose necktie, sleeves rolled up, heart earrings, wrist scrunchie</code></pre><ul><li>Anime Style:</li></ul><pre><code>anime screencap (use with the style LoRA)</code></pre><ul><li>Negative Prompts (optional):</li></ul><pre><code>dark background, blurry, rain, people
shoulder pads, detached collar, rope, sword, katana (using these may improve the main outfit)
heart (this may be necessary if you add heart earrings to the positive prompt)</code></pre>Weights Tested:<pre><code>Character LoRA: 1
Style LoRA: 0.5+ if you want her to be more recognizable
 1 if you want the anime style (with anime screencap prompt)

(Feel free to experiment with the style weight.)</code></pre>Known Issues:<ul><li>Since she always carries her sword sheath on her right shoulder while wearing her main outfit, that part of her outfit doesn't look very good. Adding shoulder pads, detached collar, rope, sword, katana in the negative prompts may help improve it.</li><li>Her hairstyle, hair color and eyes color aren't very accurate, especially when the style LoRA isn't used.</li></ul> Afterword: This time, I pruned the tags related to her hair and eyes, knowing that some people prefer shorter trigger words. However, this made these features less accurate. Maybe I should re-tag the hair and eyes of the characters in my next models... Hopefully some of you will like this model.

Files with SHA256: f4aae0e8e9e249c69c43b2f36f01d3977296bdf9523808c94289931d92e8fd63