テストが不完全です。ご注意ください。
Please note that the testing is incomplete.

1girl, solo, skirt, blue eyes, long hair, breasts, school uniform, black hair, pleated skirt, full body

トレーニング半額キャンペーンを利用してLORAにしましたが、ローカル生成環境が出来ていないのでCivitai生成です。（ちなみにテストなしの一発撮りですｗ　調整等はできていません）
あえて失敗生成も残してありますのでご参考にしてください。
とりあえず瞳の色を調整だね＾＾；
Took advantage of the half-price training campaign to create this LoRA! Since I don't have a local setup ready yet, it was generated on Civitai. (By the way, this was generated in a single attempt without any test renders lol—no prompt or parameter adjustments have been made.) I've intentionally left the failed generations as well for your reference.
First, I need to adjust the eye color ^^;

Illustrious版・Anima版と学習データは同一です。
The training data is identical to the Illustrious and Anima versions.

作成できたのが全４枚（全身はトレーニングサンプル）ですが、所感を。
とにかく最低限のキャラを維持したまま、写実とアニメ調の切り替え性能に驚きました。
欠点はありますが、テストなしの一発ですからプロンプトの工夫で救える範囲内だと思います。インテルGPUでの安定した動作方法を早く見つけなくては！という気分になりました（笑）
Here are the overall impressions based on all four generated images (the full-body shot is a training sample).
I'm genuinely surprised by how smoothly it transitions between realistic and anime styles while retaining the character's identity.
There are a few imperfections, but given that this was done on the very first try with no adjustments, it's easily fixable with prompt engineering. Now I feel an urgent need to figure out a stable workflow for Intel GPUs! (lol)