Meeting Topic: Automatic Mixing
Moderator Name: Jamie Tagg
Speaker Name: Nick Foudi
Other business or activities at the meeting: Closing semester sentiments and thanks for a great year of UML SRT AES events!
Meeting Location: Durgin Hall 209
During this meeting, second year grad student Nick Foudi gave a presentation of his Master's Thesis. His focus was on automatic mixing, whereby code he developed in Matlab has exclusive control over mix parameters from multi-track to stereo. He introduced the concept as not necessarily a way to get rid of a sound engineer's job, but to simply make it easier and faster. He described similar projects that have attempted the same goals, and pointed out some of the flaws that he wished to overcome in his own algorithms.
Describing how his program works, he first mentions that at this point, it only takes care of compression and it won't add any EQ, reverb, or effects. It decides each track's compression parameters based on a key track, which is designated based on an evaluation of all tracks dynamic range, seeking out the one with the least variation. The other tracks follow an algorithm based upon the model track that manipulated their dynamics.
Nick played back a few examples of before/after mixes with and without his program; no other processing was used. It was clear, and everyone in attendance who spoke up agreed, that the program improved the mix. In Many cases, tracks that were originally buried came through the mix well after the auto-mix processing. Also, it was apparent that the program worked better with some songs/genres than others, as each have their own general trend/style of dynamics and instrument balances.
Overall, it is a very interesting and impressive program, and leaves us wondering if auto-mixing programs will be a regular part of the industry in years to come. However, it was made clear at this meeting how complicated auto-mixing is, and where many improvements can be made in future processing schemes.
Written By: Wes Moore and Jamie Tagg