Master thesis on Sound and Music Computing Universitat Pompeu Fabra

Joaquín Jiménez-Sauma

Abstract

Finding the balance of sounds in a multitrack recording is always a time consuming process performed by experienced professionals. A poor mix produces a mix where is difficult to make the sounds stand out or enhance their presence, also creating a recording where clarity is not perceived. Auditory Masking of tracks is a common problem that affects the presence of instruments in the mix, making some elements indistinguishable and not audible.

This thesis analyses previous research in the field of automatic mixing to find methods to avoid Auditory Masking in multitrack performance. A set of tools are developed during this process to help reduce Auditory Masking by implementing different state-of-the-art techniques as well as a standard measurement of Auditory Masking. Unlike similar approaches in the state of the art, the unmasking tool implemented during this thesis is designed to work in real-time.

Previous research in the field of automatic mixing is intended to improve mixing for studio recordings. This thesis aims to apply this knowledge to the case of real-time performance where different considerations apply.

Three different versions of unmasking tools between two channels in real-time are developed as a result of this research. Each version implements different techniques with similar objectives but different advantages. The result of using this tools is evaluated both quantitatively and quantitatively to find which one provides the best results.

This repository contains the Max For Live tools developed and tested during this thesis, as well as the data from the surveys.

You can contribute to this project by answering the survey: https://jjsauma.typeform.com/to/YeLXyP

Find the thesis document on Zenodo: https://zenodo.org/record/2550903#.YDbwXy1h3T8