Lacks a bit on detail accuracy but should be flexible. Detail quality goes up when using AI upscales (hi-res fix).Suggested weight: 1This will work well at 1 with most models and while stacking most loras. You can safely try raising 0.1 at a time if issues appear that you struggle to fix with prompting. The max is quite high. You may also want to go under 1.0 but I've personally never needed to even while stacking 5+ loras.prompt suggestions:<code>ajitani hifumi, school uniform halo black pantyhose</code><code>ajitani hifumi, frilled bikini halo sunglasses</code><code>ajitani hifumi, one-piece swimsuit halo</code>less reliable<code>ajitani hifumi, playboy bunny halo [high heels:4]</code><code>ajitani hifumi, track uniform halo shorts</code>pray to RNG:<code>backpack, doll hug, stuffed toy, sneakers</code>included in training but expect garbage:<code>bullpup, grenade, holding strap</code>

[Blue Archive] Ajitani Hifumi / [ブルーアーカイブ] 阿慈谷ヒフミ

87th attempt at training this LoRa

Files with SHA256: 0703d2bc5809bd46653247d453f7b64bd005ebd8946eeef112138aff2ffbba56