Glossolalie Radio

7 channel audio installation, duration 20 minutes.
The audio that makes up Glossolalie Radio was made using two different machine learning techniques. The first is created by OpenAI, an AI research company funded by Microsoft. Their neural net software Jukebox is trained on vast datasets of music created by artists without permission. The second ML model was made using Magenta DDSP timbre transfer trained on a single voice. The mixture of these sounds are emitted from an antique radio, whilst a cloud of collaged micro-samples of the training data set plays in a hidden array of speakers in the space.
The audio is accompanied by a short text including details of Thomas Edison's belief that radio technology could communicate with the souls of the dead, the programme of a lost radio transmission from 1924 and the list of the 6012 artists used to train the OpenAI Jukebox audio model.


Love: The Revelation of Simone Weil

Audiovisual installation for Eksjö Konsthall and
Text by George Herbert


Ljudet, Rösten, Luften, Platsen

8 channel sound installation for one performer on the architecture of the voice and the plasticity of breath. Made in collaboration with Sara Wallgren and sung by Alex Peringer. 9 minute loop. Exhibited at Örebrö Konsthall and Eksjö Konsthall.


Le Congrès

Soundtrack for a film by Virginie Berthier, Easy Tiger Productions, France 3.
Text for Sonnet sur bouts-rimés by Henriette de Coligny
Performed by Arnaud Pumir, Marco Quaresimin and Isabel Sorling.


An Unheard Echo

Creation of a machine learning synthetic voice model and accompanying lament in conjunction with sculptures by the artist Sara Wallgren at Krognoshuset (Lund). The lament was played overnight on the square outside the gallery for the duration of the exhibition.

Machine learning model, recorded sound, 3 channels of audio including an outside speaker on the town square, duration 12 minutes.


Bientôt Max

Composition and voice design for the film Bientôt Max by Virginie Berthier.
Recorded audio and digital processing of voices using Max/MSP and Live, duration 20 minutes.


Into The Phone


Three Little Babes

Arrangement of a traditional melody.
Performed by Ensemble Perspectives in residency at l'Abbaye de Royaumont.