Generating lyrics with variational autoencoder and multi-modal artist embeddings
This work addresses the challenge of creating artist-specific lyrics for music generation applications, but it is incremental as it builds on existing methods with a novel initialization approach.
The authors tackled the problem of generating song lyrics conditioned on an artist's style by using a variational autoencoder with multi-modal artist embeddings, and preliminary results indicate a benefit from initializing embeddings with representations from a spectrogram classifier.
We present a system for generating song lyrics lines conditioned on the style of a specified artist. The system uses a variational autoencoder with artist embeddings. We propose the pre-training of artist embeddings with the representations learned by a CNN classifier, which is trained to predict artists based on MEL spectrograms of their song clips. This work is the first step towards combining audio and text modalities of songs for generating lyrics conditioned on the artist's style. Our preliminary results suggest that there is a benefit in initializing artists' embeddings with the representations learned by a spectrogram classifier.