An Experiment with Time

Daniele Ghisi’s An Experiment with Time (2015), is both a 3-screens multimedia installation and a live piece for ensemble, video and electronics.

This work inspired by the eponym book by John W. Dunne, depicts the writing of a diary, where the main character carries out an experiment on his own dreams, and proposes different hypothesis on the nature of time. The video loop, during 46 minutes, represents the writing of a diary during a whole year (from January to December). The starting point for the musical writing is a straightforward association between months and major chords, so that the whole year loop is handled like a sequence of perfect cadences in the tempered cycle of fifths (January being B major, February being E major, and so on, till December being F# major, and hence looping). Although the internal handling of the musical material becomes more complex (different chord types are explored and different fundamentals are used occasionally to underline specific passages), everything in the piece is conceived with respect to this simple sequence, which thus represents the skeleton of the whole musical loop.

The starting point for An Experiment with Time is the corpus of segmented audio files described. This database for is composed by about 3000 tracks of classic, contemporary, rock, pop and jazz music, sampled from the whole history of western music. The harmonic transcription of each song has been computed and a chord-based segmentation has been performed, including the four standard tonal trichords (major, minor, augmented and diminished) and a few of their basic variants. This corpus has been chosen so that time can be set a parameter of the corpus itself. The relation between the historical time and the musical time is powerful enough to create interesting diffraction patterns. As an example, during June, a radio broadcasts some sort of `history of C major’ (An officer named Major C. is also a supporting character in the video, hence the word play), composed by C major samples ordered with respect to their composition year. Similar processes are used diffusely throughout the whole work.

To compose within this database, a meta-score has been set in place, heavily relying on the bach tools.
Each note stands for the fundamental of a chord, whose type is specified via the first slot. This representation is handier than having to specify all the voices of the chord, as it allows to separately control the two orthogonal and pertinent parameters: fundamental and chord type.

For each note, additional slots carry the information for the concatenative rendering: the duration of each unit, the inter-unit distance, as well as the descriptor according to which the units should be sorted (if any) and the sorting direction. Another slot narrows the choice of units so that a certain descriptor value lies in a given range; furthermore, additional slots provide amplitude envelopes (both for each unit and for the whole sequence of units). Finally, a slot is dedicated to filtering the database according to words or parts of words appearing in the file name or path; this is an extremely quick and effective way (via Unix find command) to narrow the search to a tag or a combination of tags (e.g. `Mozart’, or `Symphony’, …).

Much more in the dedicated website:
More on the installation version of the project here.
More on the live version of the project here.
Reference: D. Ghisi, M. Bergomi, “Concatenative synthesis via chord-based segmentation for ‘An Experiment with Time'”, Proceedings of the TENOR Conference, 2016