Castamatic 4 released with Auto Leveler


The introduction of sound processing features was a turning point in the history of podcast clients. Equalization and dynamic range compression in particular can make a big difference in the listening experience for some shows or in particular listening conditions.

Dynamic range compression is a fundamental aspect of audio production, and one of the most overlooked ones when people dive into podcast production for the first time. The audio dynamic range is the difference in intensity between the loudest and the weakest signals in an audio segment. Spoken voice has a very large dynamic range (the difference in intensity between screaming and whispering is very large) and modern digital recording equipment lets us capture and reproduce it in great detail.

Our auditory system has a similarly large dinamic range, but our hearing capabilities are affected by the phenomenon of masking that occurs when the perception of one sound is affected by the presence of another, usually larger, sound. As a result of masking, audio recordings with an eccessive dynamic range are difficult to hear in many of the places where we listen to podcasts nowadays: cars, trains, sidewalks, basically any place different than a quiet room can affect the listening experience of such recordings.

That’s one of the reasons why radio stations continuously apply dynamic range compression to their output signal: when tuning to a station with a reduced dynamic range listeners can easily adjust the volume of their equipment once to an optimal level that lets them listen to the whole program in a noisy environment without the need to continuously readjust the volume knob.

Radio stations employ highly trained audio engineers, and spend great amounts of money on broadcast processors that optimize and characterize their sound. On the contrary, podcasters sometimes lack audio engineering experience and equipment, and their productions are often affected by insufficient dynamic compression. Inexpensive hardware and sub-optimal microphone positioning and technique can worsen the problem by introducing a difficult to manage noise floor and an excessively large dynamic range.

This is why dynamic range compression algorithms in podcast clients were received with great enthusiasm by the podcast listening community. The possibility of adding compression and eq while playing podcasts makes it easier to listen to less than perfectly produced shows in adverse listening conditions.

On the other side, an excessive amount of dynamic compression can adversely affect the quality of the listening experience by making music dull and unlively, rising the noise floor and producing listener fatigue.

Since version 1.0 of Castamatic the Leveler processor has tried to minimize such artifacts with careful tweaking of its time constants and the use of a particular algorithm that goes by the exotic name of gated side-chain.

Castamatic 4 goes further beyond in the direction of respecting the character of the original audio content with the introduction of a machine learning algorithm that can distinguish sections of audio that need more processing from those that are better left untouched. Auto-leveler can recognize segments that are already compressed from the ones that need processing, and most notably can distinguish music segments from spoken voice, restraining from processing the former and applying the optimal amount of processing for the latter.

The result is so transparent that I can serenely suggest to always set it to Auto. This is reflected in Auto being the default settings for new subscriptions. A Boost option is available when a lot of compression is desired; the result is quite similar to analogous options on other clients.

Castamatic 4 introduces some other notable new features:

Castamatic 4 is planned for release on June 18th 2018.