We present a case study that artistically adapts a model of emergence observed in video feedback to the design of an emergent narrative system for live performance incorporating generative artificial intelligence. We describe a methodological analogy between the creation of emergent visual patterns in video feedback and emergent narrative patterns created through narrative feedback. We explore emergent narrative through dialogical feedback loops between the improvised speech of human performers and textual summaries generated by a large language model. In this approach, the LLM is positioned as a signal path for narrative emergence rather than as a unitary source of narrative content. We encourage further artistic and theoretical exploration of narrative feedback by artists working with emergent narrative in socially interactive performance and audiovisual synthesis.