Encoding Musical Style with Transformer Autoencoders
This work addresses the challenge of controlling style and melody in music generation for AI and creative applications, representing an incremental advancement.
The paper tackles the problem of learning high-level controls over global structure in symbolic music generation by introducing a Transformer autoencoder that aggregates input encodings to capture style. It demonstrates effectiveness on music generation tasks, achieving improvements in log-likelihood and mean listening scores compared to baselines.
We consider the problem of learning high-level controls over the global structure of generated sequences, particularly in the context of symbolic music generation with complex language models. In this work, we present the Transformer autoencoder, which aggregates encodings of the input data across time to obtain a global representation of style from a given performance. We show it is possible to combine this global representation with other temporally distributed embeddings, enabling improved control over the separate aspects of performance style and melody. Empirically, we demonstrate the effectiveness of our method on various music generation tasks on the MAESTRO dataset and a YouTube dataset with 10,000+ hours of piano performances, where we achieve improvements in terms of log-likelihood and mean listening scores as compared to baselines.