Only for model inference. The training and preprocessing scripts are removed.
The XGeneration pipeline is as follows:
Given a 4-bar phrase, generate AABA-structured melody.
Download the pretrained model here.
Generate simple chord for the melody from task1, output lead sheet (melody + chord).
Generate Chinese pop style piano accompaniment based on the given lead sheet.
Download the pretrained model here.
@inproceedings{music-sketchnet,
author = {Ke Chen and Cheng-i Wang and Taylor Berg-Kirkpatrick and Shlomo Dubnov},
title = {Music SketchNet: Controllable Music Generation via Factorized Representations of Pitch and Rhythm},
booktitle = {Proceedings of the 21th International Society for Music Information Retrieval Conference, {ISMIR}},
year = {2020}
}
@article{lim2017chord,
title={Chord generation from symbolic melody using BLSTM networks},
author={Lim, Hyungui and Rhyu, Seungyeon and Lee, Kyogu},
journal={arXiv preprint arXiv:1712.01011},
year={2017}
}
@inproceedings{zhao2021accomontage,
author = {Jingwei Zhao and Gus Xia},
title = {AccoMontage: Accompaniment Arrangement via Phrase Selection and Style Transfer},
booktitle = {Proceedings of the 22nd International Society for Music Information Retrieval Conference, {ISMIR} 2021, Online, November 7-12, 2021},
pages = {833--840},
year = {2021},
url = {https://archives.ismir.net/ismir2021/paper/000104.pdf}
}