Metamorphic Efforts: Translating Kafka’s The Metamorphosis through Laban Movement Analysis.
Medialogy MSc, Embodied Interaction Project, Aalborg University Copenhagen, CREATE, Spring 2026.
A generative audiovisual translation of the opening of Franz Kafka’s The Metamorphosis, driven by Laban Movement Analysis (LMA) Effort qualities extracted through close reading of the source text. The project advances a central proposition: literary prose contains a kinesthetic body, extractable through annotation and renderable as embodied audiovisual experience.
A 10-passage walkthrough of the piece. Each passage is annotated for LMA (BESS: Body, Effort, Shape, Space) through close reading; the resulting trajectory drives a real-time TouchDesigner visual field and a five-layer polyphonic audio mix in the browser. The Action Drive arc moves Press to Wring to Glide to Slash, tracking Gregor’s progression from oppressive constriction through anguished struggle to a brief moment of release and finally to the violent break of the manager’s arrival.
Theoretical framing
The work locates the moving body in literary prose, drawing on phenomenological accounts of embodied perception (Merleau-Ponty, Dourish) and motor simulation theories of reading (Gallese and Lakoff). LMA Effort qualities (Weight, Time, Space, and Flow) serve as the descriptive vocabulary for translating textual movement cues into a generative system. Where Fdili Alaoui et al.’s pipeline goes from physical body to sensors to Effort classification, this project starts from close reading: Kafka’s text captures the corporal struggle in precise kinesthetic language, and the generative system renders the Effort qualities using the same BESS framework. Effort-to-emotion mappings draw on de Meijer’s empirical validation, and the polyphonic structure invokes Bakhtin to account for multiple simultaneous movement voices within a single passage.
Architecture
The piece is delivered through a Twine (SugarCube) browser interface that handles text, narration, and a canvas displaying live frames streamed from a headless TouchDesigner instance. TouchDesigner streams JPEG frames over WebSocket to the browser canvas, where they composite alongside five polyphonic audio layers - narration, body vocalizations, drones, sound effects, and character voices - each with independent gain control. LMA annotations reveal progressively as the reading unfolds, with hover tooltips exposing the underlying Effort qualities.
Method
Each passage is annotated for Effort qualities through close reading. The annotation is converted into a numerical trajectory and processed in JavaScript using formulas from Larboulette and Gibet (MOCO 2015) to compute live LMA descriptors. These descriptors drive an Action Drive preset-selector architecture (after Siopa et al.) in TouchDesigner, which selects from a bank of visual presets corresponding to LMA Effort drives. Character voices were generated with ElevenLabs v3; drones via ElevenLabs Music Generation; narration delivered through the browser’s audio stack.
Ongoing work
A lab study is in design to test the central proposition empirically. Participants rate movement qualities from the text alone, then experience the polyphonic piece and rate the embodied audiovisual rendering, allowing comparison of text-based and experience-based Effort ratings.
Source code and documentation: github.com/smogelmose/Embodied-interaction