<h2 id="change-log">Change log</h2><ul><li>Illustrious v3.0 The base model is Illustrious 1.0. The following masking techniques were used to make the facial style less susceptible to influence from the training data.Additionally, we reviewed the training data and adjusted the clothing colors to be closer to the original. <a target="_blank" rel="ugc" href="https://github.com/kohya-ss/sd-scripts/blob/main/docs/masked_loss_README.md">https://github.com/kohya-ss/sd-scripts/blob/main/docs/masked_loss_README.md</a> 顔のスタイルについて、学習データの影響を受けにくくしました。また、学習データを見直し、服の色調を原作に近づくように調整しました。</li><li>Illustrious V2.0 Grape The base model is Illustrious 1.0. The dataset is slightly different from v2-apple, but the creation method is almost the same. It's difficult to determine which version is superior, so this is not an upgrade. v2 Appleとデータセットがちょっと違うだけなので、アップグレードと言うよりもフレーバーの違いかと思います。好みに合う方を使ってください。 </li><li>Illustrious V2.0 Apple The base model is Illustrious 1.0. The style is more natural. And this model doesn't require the prompt "sailor senshi uniform", so this model avoids the contamination of the characteristics of characters in original work that occurs when using the prompt. </li><li>Illustrious V1.0 The base model is Illustrious 0.1. For all Illustrious and derived model. </li><li>SD1.5 V3.0 I improved the costume reproducibility of the model by further increasing the number and quality of the dataset. In addition, I increased the variety of hairstyles for characters wearing the costumes.The image number in dataset: 300(v3) &lt;- 60(v2)</li><li>SD1.5 V2.0 By renewing the images in the dataset, This Model got significant improvements in costume reproducibility and flexibility. </li></ul><h2 id="about-model">About Model</h2>This is a model of the Super Sailor Moon costume. Not Super Sailor Moon herself.It works well with character designs of any hair styles and hair colors.<hr /><h2>概要</h2>スーパーセーラームーンのコスチュームLoraです。あらゆる髪型や髪色のキャラデザインと相性が良くなるように作成しました。

Super Sailor Moon Outfit [Costume Lora] / スーパーセーラームーン衣装

I improved the reproducibility of the model by further increasing the number and quality of the dataset. In addition, I increased the variety of hairstyles for characters wearing the costumes.The image number in dataset: 300(v3) &lt;- 60(v2)

Files with SHA256: 4dec2038b06b6b82aedf05c153590b96881ba65b1725495fba5cf3bd0a989031