smart:bundle

smart:deess was designed to work on individual vocal or speech tracks and on vocal busses. As the plug-in tries to identify different phonemes in real-time, the detection quality decreases with the amount of "non voice content" in a signal. While the detection may still work for relatively simple mixes or recordings with some cross-talk or noise, smart:deess will typically have a hard time identifying individual phonemes in full mixes.

Depending on its mode, the smart:filter either only tries to fix imbalances of the current track or it also takes potential masking problems with other tracks into account (when part of a group).

In track & group (tr+gr) mode, the smart:filter looks at spectral imbalances of the current track as well as potential masking problems with other group members. This is the default mode for smart:EQ 4.

In track mode, the smart:filter only looks at spectral imbalances and problems of the current track. Potential masking problems with other group members are ignored. Even when the plug-in is a part of a group, you may want to use this mode if you do not want other tracks to have an impact on the sound of a certain track.

In group mode, the smart:filter only checks potential masking issues with other group members, but does not try to fix potential spectral imbalances of the track itself. Therefore, this mode only shows a filter curve if the plug-in is part of a group. You may want to use this mode if you want to stick to the overall sound of track and only want to embed it a bit smoother into a group of multiple tracks.

Since smart:gate detects the presence of a target source using a real-time neural network, the plug-in needs a short temporal context for the analysis. For this reason, smart:gate has a processing latency of about 80ms - which unfortunately precludes live use of the tool.

smart:limit is using the median of all measured PSR values (PSR = peak to short-term loudness ratio, see formula below) for measuring the dynamics. Although the PLR value (peak to long-term loudness ratio) is a more commonly used descriptor for dynamics, we found that our descriptor better matches the actual - also short term - dynamics of a track.

Our tests showed that the PLR value can sometimes be a bit misleading since it’s (basically) simply another descriptor for the integrated loudness. For example, if multiple tracks are normalised to -1dB True Peak, the PLR value is simply their Integrated Loudness value subtracted by 1.

PSR:

The PSR is the difference between the Short Term Loudness value and the maximum True Peak value measured within a window of 3 seconds.

PSR = Max. Peak - Short Term Loudness
=> computed within a 3 second window

PLR:

The PLR is the difference between the Integrated Loudness value and the maximum True Peak value measured over the whole observation period (e.g. typically the whole track).

PLR = Max. Peak - Integrated Loudness
=> computed over the whole observation period

The Learn function analyzes the source signal and subtly adjusts the reverb characteristics to better suit it. Unlike other sonible plug-ins, the effect here is intentionally subtle and doesn’t automatically change any parameters. Instead, it creates a customized reverb response that acts as a starting point.

In Group Mode, learning also supports cross-channel unmasking, helping reverb tails interact better across multiple instances. You can control the intensity of this influence using the Source Adaption slider.

Hint: To better hear the effect of Source Adaption, enable Wet Mode - this way you’ll listen only to the reverb signal and perceive how it’s been tailored to the source.