We introduce MusicLM, a model generating high-fidelity music from text
descriptions such as ä calming violin melody backed by a distorted guitar
riff". MusicLM casts the process of conditional music generation as a
hierarchical sequence-to-sequence modeling task, and it generates music at 24
kHz that remains consistent over several minutes. Our experiments show that
MusicLM outperforms previous systems both in audio quality and adherence to the
text description. Moreover, we demonstrate that MusicLM can be conditioned on
both text and a melody in that it can transform whistled and hummed melodies
according to the style described in a text caption. To support future research,
we publicly release MusicCaps, a dataset composed of 5.5k music-text pairs,
with rich text descriptions provided by human experts.
cite arxiv:2301.11325Comment: Supplementary material at https://google-research.github.io/seanet/musiclm/examples and https://kaggle.com/datasets/googleai/musiccaps
%0 Generic
%1 agostinelli2023musiclm
%A Agostinelli, Andrea
%A Denk, Timo I.
%A Borsos, Zalán
%A Engel, Jesse
%A Verzetti, Mauro
%A Caillon, Antoine
%A Huang, Qingqing
%A Jansen, Aren
%A Roberts, Adam
%A Tagliasacchi, Marco
%A Sharifi, Matt
%A Zeghidour, Neil
%A Frank, Christian
%D 2023
%K audio generative idea:bee_audio_llm llm
%T MusicLM: Generating Music From Text
%U http://arxiv.org/abs/2301.11325
%X We introduce MusicLM, a model generating high-fidelity music from text
descriptions such as ä calming violin melody backed by a distorted guitar
riff". MusicLM casts the process of conditional music generation as a
hierarchical sequence-to-sequence modeling task, and it generates music at 24
kHz that remains consistent over several minutes. Our experiments show that
MusicLM outperforms previous systems both in audio quality and adherence to the
text description. Moreover, we demonstrate that MusicLM can be conditioned on
both text and a melody in that it can transform whistled and hummed melodies
according to the style described in a text caption. To support future research,
we publicly release MusicCaps, a dataset composed of 5.5k music-text pairs,
with rich text descriptions provided by human experts.
@misc{agostinelli2023musiclm,
abstract = {We introduce MusicLM, a model generating high-fidelity music from text
descriptions such as "a calming violin melody backed by a distorted guitar
riff". MusicLM casts the process of conditional music generation as a
hierarchical sequence-to-sequence modeling task, and it generates music at 24
kHz that remains consistent over several minutes. Our experiments show that
MusicLM outperforms previous systems both in audio quality and adherence to the
text description. Moreover, we demonstrate that MusicLM can be conditioned on
both text and a melody in that it can transform whistled and hummed melodies
according to the style described in a text caption. To support future research,
we publicly release MusicCaps, a dataset composed of 5.5k music-text pairs,
with rich text descriptions provided by human experts.},
added-at = {2023-04-20T15:56:37.000+0200},
author = {Agostinelli, Andrea and Denk, Timo I. and Borsos, Zalán and Engel, Jesse and Verzetti, Mauro and Caillon, Antoine and Huang, Qingqing and Jansen, Aren and Roberts, Adam and Tagliasacchi, Marco and Sharifi, Matt and Zeghidour, Neil and Frank, Christian},
biburl = {https://www.bibsonomy.org/bibtex/2c64dac21a1a5d2459ca37f4b9d7efa60/annakrause},
description = {[2301.11325] MusicLM: Generating Music From Text},
interhash = {493e38be9775677f718e7367415aebec},
intrahash = {c64dac21a1a5d2459ca37f4b9d7efa60},
keywords = {audio generative idea:bee_audio_llm llm},
note = {cite arxiv:2301.11325Comment: Supplementary material at https://google-research.github.io/seanet/musiclm/examples and https://kaggle.com/datasets/googleai/musiccaps},
timestamp = {2023-04-20T15:56:55.000+0200},
title = {MusicLM: Generating Music From Text},
url = {http://arxiv.org/abs/2301.11325},
year = 2023
}