New Sonorities for Jazz Recordings: Separation and Mixing Using Deep Neural Networks.

This webpage provides sound examples related to the paper "New Sonorities for Jazz Recordings: Separation and Mixing Using Deep Neural Networks", S.I. Mimilakis, E. Cano, J. Abesser, G. Schuller, in Proc. of the 2nd AES Workshop on Intelligent Music Production, London, UK, 13 September 2016.

Source Code Available Here


Required trained models:

Trained Models (in .zip file format).



Audio Examples

Segment #1 - The Abletones Big Band - Song of India

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #2 - Dino On The Loose - Queen's Light

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #3 - The Funny Valentines - Sleigh Ride

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #4 - Jesper Buhl Trio - What Is This Thing Called Love

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #5 - Lindy Hip Big Band - The Opener

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #6 - Selwyn Jazz - Bill Ashton's 'Much Too Much'

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #7 - Zwiepack - Air

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs    


Segment #8 - Zwiepack - Summertime

Original Mixture    


Solo Source Background
Estimated Sources    
Remixed by DNNs