Synthesising Classical Guitar Pieces Using Transformer
After a literature review, no specific AI model for classical guitar in symbolic music production was found. Therefore, a new AI model was developed to generate new pieces while preserving the melodic and harmonic structure based on input notes. To do this, a dataset containing solo classical guitar pieces was tokenized using REMI and augmented. The Music Transformer deep learning model was trained on this dataset. During inference, Nucleus sampling was used to generate synthetic guitar pieces. Playability of the produced parts was evaluated with an algorithm, and a test group confirmed their pleasing qualities. With this AI, musicians, video producers can speed up their work and by listening to individualised music, conditions of people with dementia may get better.
BURAK SİNA AKBUDAK
Share this Project