Generative artificial intelligence is transforming how creativity is organised, attributed, and governed across cultural and creative industries. This study examines how generative systems redistribute creative functions within human-led workflows by comparing AI-generated outputs with live human performances across visual (VNAI) and performative (PEAI) domains. Focusing on systems such as Stable Diffusion and OpenAI’s Sora, the analysis adopts a mixed-methods approach combining semantic alignment, latent feature analysis, reverse image similarity, and motion tracking to investigate how originality, authorship, and embodied expressiveness are structured in practice. The findings show that while generative models achieve high semantic fidelity and stylistic coherence, their outputs remain constrained by training data patterns and probabilistic recombination. In contrast, live human performance exhibits greater variability, responsiveness, and affective nuance. Across both modalities, meaningful differentiation emerges through iterative prompting, selection, and contextual framing, indicating that creative value is shaped through distributed workflows in which human judgement, evaluation, and embodied interpretation play a constitutive role. These results demonstrate that generative AI systems function as augmentative components rather than autonomous creative agents. By empirically analysing how creative agency is organised across human and algorithmic processes, the study provides an analytical framework for examining authorship, labour, and value formation in AI-mediated cultural production.
Tsehaye Haidemariam (Mon,) studied this question.