PromptRefine: An Interactive Prompt Refinement Framework for Enhanced Generative AI Output

Text-to-image models such as DALL·E and Midjourney have expanded creative capabilities across various domains. However, vague or underspecified prompts often result in biased, inaccurate, or misaligned outputs. These challenges reveal a critical need for systems that better align user intent with generative model behavior. This paper introduces PromptRefine, an interactive prompt refinement framework designed to enhance clarity, specificity, and fairness in user-generated prompts. The framework consists of four modular components: a Prompt Analyzer, Clarification Generator, Refinement Loop, and Prompt Forwarder. Together, these modules guide users through an iterative clarification process that transforms ambiguous prompts into well-defined, actionable inputs. We conducted a within-subject user study with 20 participants to evaluate the system. The results showed that the refined prompts led to improved alignment with user intent, greater visual relevance, and reduced bias in the generated outputs. Participants also reported increased satisfaction and a decrease in the time required to achieve desired results. Our findings suggest that interface-level refinement tools like PromptRefine can significantly improve generative AI outcomes without requiring changes to the underlying model architectures. This offers a scalable and domain-agnostic approach to more ethical and user-aligned content generation.