lora trained on the anime girls und panzer and associated artworkmy typical approach of training a style lora and using associated tags to get characters back wasn't going to work here since many characters overlap in tags,so I took a different approach and actually had to bake in character tags. This process had to be automated so not every character was captured and there may be mistakes that might show up in generation . Some prompt coaxing also might still be necessary. suggested base checkpoint to use it 7th animetrigger words :<pre><code>akiyama_yukari
anchovy_(girls_und_panzer)
darjeeling_(girls_und_panzer)
itsumi_erika
nishizumi_maho
nishizumi_miho
katyusha_(girls_und_panzer)
kay_(girls_und_panzer)
mika_(girls_und_panzer)
takebe_saori
pepperoni_(girls_und_panzer)
reizei_mako
shimada_arisu
isuzu_hana
kadotani_anzu
kawashima_momo
kondou_taeko
marie_(girls_und_panzer)
orange_pekoe_(girls_und_panzer)
rosehip_(girls_und_panzer)
andou_(girls_und_panzer)
assam_(girls_und_panzer)
hoshino_(girls_und_panzer)
nishi_kinuyo
nishizumi_shiho
nonna_(girls_und_panzer)
sawa_azusa
cutlass_(girls_und_panzer)
erwin_(girls_und_panzer)
</code></pre>

Files with SHA256: 2a486f7187fd295f54cc3f921bddf5c5ff30d4e5451f00fc0c4d245fe5e01cb1