Abstract

This research investigates the philosophical and technical implications of applying generative artificial intelligence to film restoration, focusing on the notion of plausibility as a new aesthetic and epistemological paradigm. The study explores how AI-based frame interpolation models can reconstruct missing or degraded sequences in early twentieth-century films, not by recovering lost information, but by generating perceptually coherent continuities between surviving frames. In this context, authenticity no longer coincides with material originality but with the consistency of visual experience. To assess the boundaries and potential of interpolation-based restoration, we introduce a three-tier framework comprising Frame Complexity, Motion Complexity, and Damage Complexity (and an additional one Negative Frame Complexity). A case study on The Rival Sculptors (1911) demonstrates how plausibility can operate as a guiding principle for preserving cinematic experience across time and technology. By bridging aesthetic theory and computational methodology, the research suggests that AI does not simply reproduce the past but actively constructs believable futures for visual memory.