In this case I've experimented with prompts generated by LLM. Mostly it was Qwen 3.5 9B uncensored by HauhauCS. Sometimes generative model can understand very unusual prompts with interesting results.
Funny to see a spinoff of post #148179 but why not bother to remove unnecessary model and meta tags like "chatgpt, google gemini" from the prompt? She also seems implied to be staring at the unseen wall.
