Looks like a Steinberg Free Filter rip-off to me!
Not anymore! Har-Bal 2.3 did similar things to the Free Filter plugin but with significant differences. Har-Bal 3 does things Free Filter cannot do : track splitting and complex filters, mid-side processing and spatialising Har-Bal air just to name a few differences.
Har-Bal is stand-alone which affords it a degree of usability that is not possible with a plugin.
The way you design an EQ filter in Har-Bal is totally different and more flexible. Free Filter is constrained to 1/3 octave. Har-Bal has no such constraint.
The way Steinberg’s Free Filter and Har-Bal assess the spectral content of the music is very different too. Har-Bal works off the average of the entire track because the average of the entire track will best describe peaks that will cause listening fatigue. Looking at an RTA you get some idea but it isn’t as clear cut as an average for the entire track. Har-Bal also tracks the peak spectrum too.
The biggest difference as far as usability is concerned is compensating for perceived loudness effects which, as far as we’re aware, only Har-Bal does. What do we mean by that?
A simple example can be visualized as follows.
If you had a graphic EQ in front of you and you realized a particular frequency using only boost and on a parallel one you implemented the same frequency response using only cut. Now do an A/B test on both and which sounds better? Obviously the one that uses only boost because it is louder, although from an Eq’ing point of view they are identical!
You might say this is a trivial case and you should adjust the EQ to use nominally the same boost and cut and then you’d only have one “correct” frequency response. Wrong! This would only be true if the source material was largely “pink” and human hearing wasn’t frequency and level dependent! In performing mastering equalization neither of these assumptions are true so how do you account for the different perceived loudness in different EQ arrangements when performing A/B assessments? At worst you don’t and at best you guess. Neither approach is objective.
Har-Bal takes an educated guess which is objective and likely to be far less biased than any manual approach. It does so by calculating a loudness index for the pre and post equalization spectrums and adding or removing the appropriate gain to the EQ’d case to remove loudness related biases. It is an educated guess in the sense that the Zwicker loudness measure is only approximately indicative of human loudness perception but it’s a far better approximation than being left to your own devices.