Decoupling Performance and Effects in Virtual Production via AI-driven Post-Production

To address the structural conflict in Virtual Production (VP) between real-time rendering pipelines and the demands of nonlinear performance and complex visual effects creation, this paper proposes a Lighting-Guided Generative Workflow (LGGW). The core of this approach lies in temporal decoupling: controllable DMX-based physical lighting is used as both an on-set interaction medium and a post-generation anchor, allowing high-risk visual effects to be detached from real-time rendering. Empirical evidence from the sci-fi short film sequence “Meteor Fall” demonstrates that this workflow enables generative AI to asynchronously complete footage embedded with lighting cues, achieving pixel-level integration and compressing production time from days to hours. By preserving lighting coherence and performance authenticity, LGGW introduces a new paradigm for cost-effective hybrid production that integrates physical and virtual elements.