[image 03694] Sources for AI Melody Generation from Lyrics
yiyu ＠ nii.ac.jp
2019年 8月 30日 (金) 15:28:47 JST
Dear colleagues in image forum,
This is Yi Yu from NII (http://research.nii.ac.jp/~yiyu/ ).
(I would like to appreciate your kindness if you kindly forward my email to colleagues who are interested in AI melody generation from lyrics.)
I would like to bring your attention to our recent research on conditional LSTM-GAN for melody generation from lyrics at https://arxiv.org/abs/1908.05551 .
Sources for Neural Melody Generation from Lyrics can be found at
What can you benefit from these sources?
1. A big music dataset of MIDI files with the alignment between lyrics and melody (providing IDs of original MIDI songs and lyrics embedding vectors related to syllable-level, word-level, and sentence-level)
2. Lyrics encoder model trained by big lyrics dataset (based on syllable-level, word-level, and sentence-level). You can directly use it as a tool to extract lyrics embedding vector.
What kinds of research can you try (more than what I suggest)?
1. Neural melody generation from Lyrics
2. Neural lyrics generation from melody
3. Compose melody with sketch of incomplete lyrics
4. Compose lyrics with incomplete melody
5. Compose your own lyrics and melody for video or image
6. Generate melody for video based on moods