PRISM offers a revolutionary black-box approach to automated prompt generation, improving versatility and interpretation across T2I models. Inspired by LLM jailbreaking, it uses the in-context learning ability of LLMs to enhance prompts with human-understandable candidates, validated by its effectiveness across different models and concepts.

• Novel algorithm for automated prompt engineering
• Human-interpretable and transferable prompts generation
• Black-box access and in-context LLM learning utilization

As T2I generative models become central to creative industries, automated and intuitive prompt engineering like PRISM could significantly streamline content creation. Its adaptability across models underscores the potential for widespread adoption and future innovations in automated artistry and design.