Linguitone: a web-based integrated system to support the creation of electric guitar sound effects using natural language expressions

Atsuya Kobayashi’s Graduate Thesis Project

In electric guitar performance, when “sound creation” is performed by changing the sound with effect pedals, unique linguistic expressions are used to share the sound image among guitarists. However, beginners do not connect the linguistic expression of the sound image with the actual sound image, and it is difficult for them to create a concrete sound image from these tone expressions and to actually create a sound using an effect pedal. In this study, we focused on the patterns of linguistic expressions in timbre phrases, and developed Linguitone, a system that supports sound creation using a machine learning model that predicts parameters of guitar effectors from embedding vectors of natural language. We evaluated the usability and supportability of the system with electric guitar players.