<h1 id="rappa-()-from-honkai:-star-rail-ppuxdxpeg">Rappa (乱破) from Honkai: Star Rail</h1> A peculiar girl who appears in Penacony like a flashbang at the darkest hour of night, identifying herself as a ninja and attributing everything in the world to "ninjutsu." Upholding the recitation of ninja mantra, creating Dazzling Ninja Seals, and mastering ninja techniques — That is, the Way of the Ninja involves rap, graffiti, and comics — Through rigorous self-discipline, she roams the stars, upholding justice and righteousness. As a member of the Galaxy Rangers, she relentlessly pursues the villain known as Evil Ninja Osaru, chasing them to the very edge of the Cosmos.Since she's nearly here, I decided to update this LoRA as per my promise since there's more images now. Not a lot, but still enough. I've also upscaled the dataset so that the images you generate are more high quality.Adetailer is recommended.<h2 id="weight-p25zwksv8">Weight</h2>I recommend using a weight between 0.8 and 0.1. If you can't decide, I find that 0.9 works best.<h2 id="trigger-words-qivobhf72">Trigger Words</h2><h3 id="base-6kllgwxtp">Base</h3><pre><code>rappa_hsr, pink hair, ponytail, long hair, green eyes, hat, baseball cap, fake horns</code></pre>You can remove her hat by removing "hat, baseball cap, fake horns" in the prompt.<h3 id="clothes-kp3vjnfw4">Clothes</h3><pre><code>red jacket, open jacket, tied shirt, midriff, fingerless gloves, navel, black shorts, short shorts, belt, black footwear, toeless footwear</code></pre><h2 id="stats-for-nerds-9q9i6nxai">Stats for Nerds</h2>v2Trained with <a target="_blank" rel="ugc" href="https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb">hollowstrawberry's XL Lora Trainer</a><ul><li>Images in dataset: 52</li><li>Repeats: 6</li><li>Epochs: 10</li><li>Training steps: 800</li><li>Unet: 1e-4</li><li>Text_encoder: 1e-6</li><li>Dim/alpha: 8/4</li><li>Optimizer: AdamW8bit</li></ul>v1Trained with <a target="_blank" rel="ugc" href="https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb">hollowstrawberry's XL Lora Trainer</a><ul><li>Images in dataset: 29</li><li>Repeats: 10</li><li>Epochs: 10</li><li>Training steps: 750</li><li>Unet: 5e-4</li><li>Text_encoder: 1e-4</li><li>Dim/alpha: 16/8</li><li>Optimizer: Prodigy</li></ul>

Rappa | 乱破 (Honkai: Star Rail)

<ul><li>Added more dataset images (they're also upscaled)</li><li>Changed recommended tags</li><li>Filesize is a lot smaller now</li></ul>

Rappa_HSR-v2.safetensors

v2.0

v1.0

<h1 id="rappa-()-from-honkai:-star-rail-ppuxdxpeg">Rappa (乱破) from Honkai: Star Rail</h1> A peculiar girl who appears in Penacony like a flashbang at the darkest hour of night, identifying herself as a ninja and attributing everything in the world to "ninjutsu." Upholding the recitation of ninja mantra, creating Dazzling Ninja Seals, and mastering ninja techniques — That is, the Way of the Ninja involves rap, graffiti, and comics — Through rigorous self-discipline, she roams the stars, upholding justice and righteousness. As a member of the Galaxy Rangers, she relentlessly pursues the villain known as Evil Ninja Osaru, chasing them to the very edge of the Cosmos.Since she's nearly here, I decided to update this LoRA as per my promise since there's more images now. Not a lot, but still enough. I've also upscaled the dataset so that the images you generate are more high quality.Adetailer is recommended.<h2 id="weight-p25zwksv8">Weight</h2>I recommend using a weight between 0.8 and 0.1. If you can't decide, I find that 0.9 works best.<h2 id="trigger-words-qivobhf72">Trigger Words</h2><h3 id="base-6kllgwxtp">Base</h3><pre><code>rappa_hsr, pink hair, ponytail, long hair, green eyes, hat, baseball cap, fake horns</code></pre>You can remove her hat by removing "hat, baseball cap, fake horns" in the prompt.<h3 id="clothes-kp3vjnfw4">Clothes</h3><pre><code>red jacket, open jacket, tied shirt, midriff, fingerless gloves, navel, black shorts, short shorts, belt, black footwear, toeless footwear</code></pre><h2 id="stats-for-nerds-9q9i6nxai">Stats for Nerds</h2>v2Trained with <a href="https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb" rel="ugc noopener noreferrer nofollow" target="_blank">hollowstrawberry's XL Lora Trainer</a><ul><li>Images in dataset: 52</li><li>Repeats: 6</li><li>Epochs: 10</li><li>Training steps: 800</li><li>Unet: 1e-4</li><li>Text_encoder: 1e-6</li><li>Dim/alpha: 8/4</li><li>Optimizer: AdamW8bit</li></ul>v1Trained with <a href="https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb" rel="ugc noopener noreferrer nofollow" target="_blank">hollowstrawberry's XL Lora Trainer</a><ul><li>Images in dataset: 29</li><li>Repeats: 10</li><li>Epochs: 10</li><li>Training steps: 750</li><li>Unet: 5e-4</li><li>Text_encoder: 1e-4</li><li>Dim/alpha: 16/8</li><li>Optimizer: Prodigy</li></ul>

&lt;ul&gt;&lt;li&gt;&lt;p&gt;Added more dataset images (they&#39;re also upscaled)&lt;/p&gt;&lt;/li&gt;&lt;li&gt;&lt;p&gt;Changed recommended tags&lt;/p&gt;&lt;/li&gt;&lt;li&gt;&lt;p&gt;Filesize is a lot smaller now&lt;/p&gt;&lt;/li&gt;&lt;/ul&gt;

Files with SHA256: 65c8b6b5534e8f3c59e05cc1b8260350560def3ab0950fda677ea686b3994c42