Towards Holistic Prompt Craft

2025-04-08

AI Summary

This study addresses the inefficiency in real-time AI image generation caused by the fragmentation among prompts, parameters, and interface feedback. To resolve this, we propose “holistic prompt engineering” and introduce the Holistic Prompt Craft framework. We designed two interactive prototypes—PromptJ (a prompt mixer) and PromptTank—and conducted three field-based art experiments, reflective design workshops, and conceptual abstraction to enable concurrent, integrated control over generative elements. Our contributions include: (1) the first formal articulation of the “strong-concept” methodology for prompt design; (2) the establishment of a reusable, intermediate-level knowledge system bridging abstract principles and implementation; and (3) a theoretically grounded yet practice-oriented framework for human-AI co-creation with generative AI. This work advances prompt paradigms from tool-centric operation toward human-centered, real-time collaborative ideation.

Technology Category

• Natural Language Processing: Prompt Engineering / Prompting
• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

We present an account of an ongoing practice-based Design Research programme that explores the interaction affordances of real-time AI image generators. Based on our experiences from three installations, we reflect on the design of PromptJ, a user interface built around the concept of a prompt mixer. Our first contribution is a series of strong concepts based on our reflections of designing and deploying PromptJ. We cohere and abstract our strong concepts into the notion of Holistic Prompt Craft, which describes the importance of considering all relevant parameters concurrently. Finally, we present PromptTank, a prototype design which exemplifies the principles of Holistic Prompt Craft. Our contributions are articulated as strong concepts or intermediate knowledge that are intended to inform and inspire practitioners and researchers who are designing with image generation models or developing novel interaction paradigms for generative AI systems more generally.

Problem

• Exploring interaction affordances of real-time AI image generators
• Designing PromptJ interface around prompt mixer concept
• Developing Holistic Prompt Craft for concurrent parameter consideration

Innovation

• Real-time AI image generators interaction affordances
• PromptJ UI with prompt mixer concept
• PromptTank prototype exemplifying Holistic Prompt Craft