This is an experimental style lora intended to produce a semi-realistic digital concept art style. It is trained entirely on artificial data from a Z-Image style that I found pleasing. I've found that for whatever reason, certain prompts will cause sudden anime style. There's no anime in the dataset! This is some strange quirk of Moss. I'd recommend putting 'anime' in the negatives. There is no trigger word required, but there are some things that can helpHelpful words: traditional oil painting style, semi-realism,

Files with SHA256: dac78fa961a85ee0aa383a51652ea4ab2cd96b0d1bff4e86de83b1d48e1810c1