Preconstructed aura in generative synthesis a discursive reading of memory

Abstract

This theoretical essay investigates, from the standpoint of Discourse Analysis, what happens to the notion of aura and to its mnemonic grounding when images, videos and sounds come to be born from utterances, natural-language commands sent to generative models. Starting from Walter Benjamin and a domestic scene, it reads the aura as a relation of memory and follows it through three shifts: from the tradition of groups to the technical archive, from the archive to the image, and from the image back to collective memory. Drawing on Michel Pêcheux’s notions of discursive memory and the preconstructed, Eni Orlandi’s metallic memory and silence, Maurice Halbwachs’s collective memory and Paul Ricoeur’s distinction between memory and imagination, it argues that generative synthesis neither eliminates the aura nor transfers it to whoever writes the command, but alters the supports, scales and rhythms of its attribution. Images reach groups with an effect of uniqueness calculated from databases, an already seen aspect that presents itself as unique, here called preconstructed aura, and the command works as a caption written before the image and as a commentary on an archive restricted to a few companies. The value that makes these images unique, however, still depends on the social frameworks of memory, in which groups share, validate or reject them, and it is in this coexistence that a generated image may become the memory of an event that never happened.