The dataset is made of anime screencaps.I didn't train specific outfits, as there aren't really any. Sadly, i couldn't get the eyes to work, as they were a pretty subtle detail. The tag "mbase" might help with accuracy of the character. Otherwise, stick to the tags I have listed on CivitAI. I just used the parameters i was using for PonyXL LoRAs. Please tell me if there's anything wrong (errors, artifacts, etc.), so i can try avoiding those in the future.

Mio - Tsuki ga Michibiku Isekai Douchuu (TSUKIMICHI -Moonlit Fantasy-)

An epoch with relatively good balance between likeliness to the original and overall quality.Using even later epochs shouldn't change anything.

MLF-Mio_ILL_epoch_12.safetensors

[ILL] Epoch-12

Creator: <a href="https://civitai.com/user/nrocka" rel="ugc noopener noreferrer nofollow" target="_blank">Nrocka</a>Base triggers:<ul><li>mio-mlf, black hair, short hair, blunt bangs, bob cut, bangs, yellow eyes</li><li>mbase</li></ul>1 Outfit:<ul><li>japanese clothes, kimono, bow, black gloves</li></ul>The tag "mbase" might help with accuracy of the characterUsing tags "anime screencap", or just "screencap" might help though

Files with SHA256: 66c528c672e5496989bd566e6fc0cf9d889fde0d2868d020b93292fc05994a14