Tormenting the Generating AI
tare_koala
127 Followers·615 Posts
I'm writing prompts to generate accurate and precise (and hopefully even cuter) models of Japanese women's uniforms and attire. These are generally publicly available, so please feel free to take them (at your own risk). Be aware that many are long models for the latest styles. Compliments on my posts are welcome, as are any opinions or questions. *I don't dislike R15 content, but it's not my main purpose. I mute accounts with many images of underwear, large breasts, or extreme swimsuits, and in severe cases, I block them. Thank you for your understanding.
At 175 tokens, it's not particularly long, but it's a challenging prompt for image generating AI. We'd appreciate it if you could test it with your usual model. There must be no inconsistencies between the girl reflected in the mirror and her back, which is not reflected in the mirror. Furthermore, it's forced to generate a back view in response to a prompt that usually assumes a front view. Even with the latest model, gpt-image-2, there's usually a mistake somewhere, as seen in this instance where the left hand is inconsistent. Since it's a rear view, using slightly older models like flux/schnell might result in the ribbon being on the back. Endpoint: fal-ai/gpt-image-2, quality: low, image_size: 1024x1536, Duration: 22.427247s, Cost: $0.0057=0.905yen/1piece
AI evaluation
Great
Strong compositionStrong color
AI
鏡像を通じて自己と向き合う少女の姿は、秋の紅葉が織りなす儚い情景と相まって、内省的な美しさを醸し出している。構図の対称性と色彩のコントラストは巧みだが、全体の調和や独創性においてさらなる深みが求められる作品である。

