Audio-to-Lyrics Alignment

This the audio-to-lyrics alignment system I published in ICASSP2021. The cool thing about this is that it works on music recordings with any duration. It first applies a biased search to find anchoring word timings within the recording, then applies alignment in between anchoring words.
A demo can be found on the main page.
The code and the pretrained model can be found here

Next

ALTA: Training recipe for the state-of-the-art lyrics transcriber

This repository contains the complete trainining pipeline for the MStreNet based Automatic Lyrics Transcription module. Key features: Compact Multistream TDNN, Cross-domain Training, Music-informed silence modeling

Next

JAAH: Audio-aligned Jazz Harmony Dataset

Next

Automatic Makam Recognition

Support Vector Machine-based automatic Makam recognition. The system can tell you which (dominant) makam type your input song is played on. We used the Ottoman-Turkish Makam Music dataset for training the classifier. Work in collaboration with Baris Bozkurt and Xavier Serra.

The code can be found at here.

Next