Hector Rodriguez

Rhuthmus (2024) is a computational art project that uses machine learning to compare and contrast moving image sequences with respect to their visual rhythms.    Although essential to cinematic art, visual rhythm is difficult to define and characterize. The temporal fluidity of images resists linguistic categorization. The question arises as to whether machine learning can potentially capture or articulate certain aspects of cinematic experience which would otherwise remain elusive.    The main element in the Ruthmus project is a neural network trained to represent certain aspects of cinematic rhythms. This representation can be visualized by selecting pairs of short sequences that, according to the system, exhibit similar rhythms. If the template includes a cut, as in the last section of the video, the system returns matches which also include cuts. The system can potentially be used to capture editing patterns in a movie.    The system has not been trained to recognize faces, but when the template is a face, the system tends to return faces, which might suggest that there are rhythms roughly characteristic of faces.   The current hype around generative AI emphasizes the power of machine learning to “hallucinate” new images and text. This fascination with generativity overlooks technology's potential to enhance our understanding of existing artworks. Machine learning can help film scholars to discern, attend to, and describe visual patterns in classical movies.        Link to documentation:  https://concept-script.com/rhuthmus/main.html

Credits:

ARTISTIC and TECHNICAL DIRECTOR

Héctor Rodríguez

PROGRAMMERS

Mou Peijing

Héctor Rodríguez

EDITING and WEB DESIGN

Mou Peijing

PRODUCER

Richard Allen