Test 11.
Trigger word: Yuj1N1reh474
(And "traditional media" will significantly affect the final image result.)
The LoRa model for artist Yuji Nirehata (楡畑雄二), not cartoonish, it is a retro Japanese adult magazine illustration style. All images used in the training were black and white illustrations.
This model was released at the request of a friend. I know very little about this artist, but I found that this artist has a large number of works in the series titled "Shy Lewd Lock (羞弄の淫鎖)" and the style is somewhat different from his other works.
Training by mixing all the images together didn't yield good results. So I eventually separated the "Shy Lewd Lock" series of artwork from the artist's other works, and trained each of them into an independent LoRa model.
The LoRa for the "Shy Lewd Lock" style is named "lock", and the LoRa for the other styles are named "lite".
Ver.2 isn't necessarily better than Ver.1. V1 used more images and manually added tags. V2 used fewer images, but clearer ones, and tags were added automatically by AI. V2 seems to reproduce the art style better, but it doesn't perform as well as V1 when generating certain specific content.
These two loras can be used together. They have the same keywords, but the strengths need to be reduced, and the sum of their strengths should not exceed 1.