VIDEO DESCRIPTION:

Play, download and edit the free video Clamabits.

I have had a long-standing fascination with the spatialisation and vocal gestures employed in performances of Monteverdi's Vespers of 1610, and the tenor duet (actually employing a third tenor as a distant echo) is a remarkable piece with some virtuosic passages. This is way beyond my capability, but I couldn't resist trying a feasibility study of this as the final project for Kadenze's Machine Learning for Musicians' course.

I have used two fragments of Duo Seraphim and use syllable-level recognition to trigger different three sound treatments: a duet harmoniser, a record and playback with segmentation finished with a trio harmonizer, consistent with the text of the piece.

Although I originally set this up for six syllables, the final two "tres" and "sunt" registered so many false positives, that I decided not to use them.

The most important part for me was to start, stop and playback a recording in real time, and this project has shown me how to achieve that.

Overall, I think this is quite a fragile setup: it is hard to discriminate between most of the gestures, and I would not want to rely on it in a performance. I think this is largely due to my choice of gestures, and the fact that the piece is predominantly vowel sounds.

Download, play and edit free videos and free audios from Clamabits using RedcoolMedia.net web apps