# vocalMashup

**Repository Path**: segmentationFaults/vocalMashup

## Basic Information

- **Project Name**: vocalMashup
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-08-11
- **Last Updated**: 2021-11-02

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# vocalMashup
Creates a mashup from a given vocal recording and a list of background instrumental tracks. 

NOTE : The mashup pipeline is very rudimentary. If you use better algorithms for tempo detection, key detection, the result will be much better :) 

### How it works 
Computes tempo, beat and key from vocal and background tracks to find the best match.   
All the algorithms are from [librosa](https://librosa.github.io/librosa/).  

### How to use 
I added some test audios from online and [musdb18](https://sigsep.github.io/datasets/musdb.html).   
1. Put background tracks you want to mash into `background_tracks` directory. 
2. Run the following with a vocal track to mash
```
python mashability.py test_vocal.wav
```

### To improve mashup quality
* Better algorithms for feature computations (try with [madmom](https://github.com/CPJKU/madmom))
* Better mixing technique


### Projects using this code
* "Learning a Joint Embedding Space of Monophonic and Mixed Music Signals for Singing Voice" - Kyungyun Lee, Juhan Nam, ISMIR 2019 ([pdf](https://arxiv.org/abs/1906.11139), [github](https://github.com/kyungyunlee/mono2mixed-singer/))