This Lora was trained using 212 images from the official Blue Archive series, season 1 with 12 episodes.No fan art or AI-generated images were used; I wanted to keep the art style as close to the original as possible.I checked all the images, removing the worst or blurry ones, as well as other characters in the background to avoid training errors.Based on my quick tests, a weight of 0.8 is ideal, but it may vary from one checkpoint to another, so feel free to try different weights.I did not use trigger words in the Lora, it was trained with a tag database, however I advise you to use<pre><code>hoshino (blue archive),</code></pre>Here is my test setup (note that this isn't necessarily the most optimized) :<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1f100faf-30a6-4ec4-ad11-43828e893455/original=true/1f100faf-30a6-4ec4-ad11-43828e893455.jpeg" />I'm using the official Anima Base checkpoint version 1.0 : <a target="_blank" rel="ugc" href="https://civarchive.com/models/2458426/anima?modelVersionId=2945208">Link.</a>

Files with SHA256: 59da3eb77f016dd2b4ca22322a30e2e3918ea691e005778fc7891d7c27a427c6