The Machine Sees, the Machine Tells? Montage Theory and Generative Narratives in the Age of AI

Source of Publication

Proceedings of the 31st International Symposium on Electronic/Emerging Art (ISEA2026)

Conference Year

2026

Abstract

Montage can be reframed as a shared epistemological principle linking Soviet film theory and machine learning and generative artificial intelligence. Drawing on Eisenstein’s dialectical collision of shots, Kuleshov’s experiments on contextual emotion, and Vertov’s kino-eye, we argue that cinematic meaning is produced relationally between images, across time, and through viewer inference, rather than residing in isolated frames. We parallel this with AI systems that learn through statistical relationships, then contrast frame-based recognition with temporally aware video transformers, generative diffusion models, and multimodal models that align text and image as a cross-modal montage. Against the assumption that training on finished films yields narrative understanding, we argue that cinematic reasoning is distributed across intermediate forms—scripts, storyboards, rehearsals, and editorial decisions, and that practice-based pipelines using structured representations can improve coherence and creative control. AI is therefore positioned less as a decoder of cinematic meaning than as a speculative montage partner that proposes rhythms, juxtapositions, and associations for artists to interpret and contest. We extend montage beyond cinema to short-form, algorithmically sequenced media, where rapid cuts and repetitions generate affect and political interpretation through relational cues. The conclusion calls for renewed media literacy and agency over machine-made connections in everyday visual culture.

Document Type

Conference Proceeding

Publication Date

11-6-2026

This document is currently not available here.

Share

COinS