Movement Invisible

Movement Invisible is an experiment on movement and artificial intelligence, an exploration on the relation between the body and its surroundings, both in the physical world and the digital one. What mechanics govern our behaviour in these domains and can they be made visible through dance? Invisible Movement explores the underlying mechanics by having dancers interact with an algorithm on stage. This is done through the combination of tracking technology, projection mapping and machine learning systems.

Three different machine learning systems are combined to establish the explorative space for the dancers. A console enables the dancers to interact directly with the machine, entering text prompts that are processed by a text2text and a text2image system. The textual and visual responses are projected back onto the space, on the surrounding walls and the floor. A third system tracks the movement of a dancer and used as a 3D drawing mechanic, creating visuals based on their skeleton postures.

The language that the dancers use range from minimal movements and transits through space to more developed material originated from hidden movements present in the body. Movement generated from body parts that are inhibited and forgotten. This is researched through improvisation, which plays an important part in the experience.

Technology

The infrastructure was built up around realtime socket connections, to facilitate fast response times on control changes and integration of tracking data. At the core is a realtime server that all components connect to. Through a web interface and connected midi-controller the dancer can interact with the 3 machine learning systems. The visual answers are send through a webpage to multiple projectors, one for each surface that the dancer can chose during interaction.

Separate clients were written in Python that act as brokers to the server, one for the text2text and text2image ML systems and one for the tracking data.

High level overview of the Movement Invisible AI architecture based on three primary components: the realtime control server, the AI model and tracking processor, all connected through realtime socket connections for fast response times. An independent number of projections can be connected up.

Performances

Das Research program Queer Feedback Sessions Spring 2023 / Academie voor Theater en Dans, Amsterdam / 01-04-2023
Das Research program Queer Feedback Sessions Spring 2023 / DAS Graduate School, Amsterdam / 11-02-2023
Come Together 8 festival / BAU Dance & Performance Amsterdam / Frascati Theater, Amsterdam / 19-01-2023

Credits

Movement Invisible is a project of choreographer and dancer Valentina Campora, interdisciplinary artist Luciano Pinna and musician and producer Andy Moor.

With:
Valentina Campora, Kim Hoogterp and Liah Frank: Dancers
Andy Moor: Musician
Luciano Pinna: Photography and AI operator

Movement Invisible uses the text2image model based on Stable Diffusion-v1-4 and the textgen GPT-Neo 1.3B transformer model is designed using EleutherAI.

For more photos about the experience in Frascati Theater during the Come Together 8 festival, click here. And more on the performance in Academie voor Theater en Dans as part of the Das Research program Queer Feedback Sessions Spring 2023, click here.

Frascati Theater performance
Academie voor Theater en Dans performance