Adaptive Accompaniment with ReaLchords

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Wu, Yusong, Cooijmans, Tim, Kastner, Kyle, Roberts, Adam, Simon, Ian, Scarlatos, Alexander, Donahue, Chris, Tarakajian, Cassie, Omidshafiei, Shayegan, Courville, Aaron, Castro, Pablo Samuel, Jaques, Natasha, Huang, Cheng-Zhi Anna
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908410861060096
author Wu, Yusong
Cooijmans, Tim
Kastner, Kyle
Roberts, Adam
Simon, Ian
Scarlatos, Alexander
Donahue, Chris
Tarakajian, Cassie
Omidshafiei, Shayegan
Courville, Aaron
Castro, Pablo Samuel
Jaques, Natasha
Huang, Cheng-Zhi Anna
author_facet Wu, Yusong
Cooijmans, Tim
Kastner, Kyle
Roberts, Adam
Simon, Ian
Scarlatos, Alexander
Donahue, Chris
Tarakajian, Cassie
Omidshafiei, Shayegan
Courville, Aaron
Castro, Pablo Samuel
Jaques, Natasha
Huang, Cheng-Zhi Anna
contents Jamming requires coordination, anticipation, and collaborative creativity between musicians. Current generative models of music produce expressive output but are not able to generate in an \emph{online} manner, meaning simultaneously with other musicians (human or otherwise). We propose ReaLchords, an online generative model for improvising chord accompaniment to user melody. We start with an online model pretrained by maximum likelihood, and use reinforcement learning to finetune the model for online use. The finetuning objective leverages both a novel reward model that provides feedback on both harmonic and temporal coherency between melody and chord, and a divergence term that implements a novel type of distillation from a teacher model that can see the future melody. Through quantitative experiments and listening tests, we demonstrate that the resulting model adapts well to unfamiliar input and produce fitting accompaniment. ReaLchords opens the door to live jamming, as well as simultaneous co-creation in other modalities.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14723
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adaptive Accompaniment with ReaLchords
Wu, Yusong
Cooijmans, Tim
Kastner, Kyle
Roberts, Adam
Simon, Ian
Scarlatos, Alexander
Donahue, Chris
Tarakajian, Cassie
Omidshafiei, Shayegan
Courville, Aaron
Castro, Pablo Samuel
Jaques, Natasha
Huang, Cheng-Zhi Anna
Sound
Artificial Intelligence
Jamming requires coordination, anticipation, and collaborative creativity between musicians. Current generative models of music produce expressive output but are not able to generate in an \emph{online} manner, meaning simultaneously with other musicians (human or otherwise). We propose ReaLchords, an online generative model for improvising chord accompaniment to user melody. We start with an online model pretrained by maximum likelihood, and use reinforcement learning to finetune the model for online use. The finetuning objective leverages both a novel reward model that provides feedback on both harmonic and temporal coherency between melody and chord, and a divergence term that implements a novel type of distillation from a teacher model that can see the future melody. Through quantitative experiments and listening tests, we demonstrate that the resulting model adapts well to unfamiliar input and produce fitting accompaniment. ReaLchords opens the door to live jamming, as well as simultaneous co-creation in other modalities.
title Adaptive Accompaniment with ReaLchords
topic Sound
Artificial Intelligence
url https://arxiv.org/abs/2506.14723