
This model is inspired by the color lithographic works of Oda Kazuma from the early 20th century. Unlike his more famous ukiyo-e (woodblock) artwork, his lithographs gave off a much more Western impressionistic artstyle, which makes sense given his studies under an Austrian lithographer.
Keyword:
odkzsb
Negative prompts:
shin-hanga, photograph
Kazuma often drew landscapes and everyday scenes, which is where this model shines. However, it can also do portraits and other things. When making portraits, I usually add photograph
as a negative prompt, and/or increase the weight from 1.0 up to 1.25, in order retain the drawn style. Kazuma drew most women in a traditional Japanese style, so Asian women may skew towards that look as well.
Additionally, since his work is quite similar to woodblock printing, adding shin-hanga
as a negative prompt is necessary!
All of the source photos came from the NDL public domain gallery on Oda Kazuma, of which you can find some here (it's fun to compare!); However, his monochrome (often sepia toned) lithographs will eventually be adapted into a separate LoRa because they are visually distinct and different from his color lithographs.
Note: Do not use the SDXL Refiner with this LoRa, unless you are okay with it removing the details unique to the artist's style.
描述:
Drastically improved the stylistic resemblance to the source material without overfitting to the source material. Unlike 0.6, which was more abstract, this version aligns far more closely to the intended style. I went through about 22 trainings to get this result.
Style weights and keywords/negative prompts have stayed the same.
训练词语: odkzsb
名称: odkzsb_pro_v28_2400.safetensors
大小 (KB): 104207
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success