Tom's Audio Processing plugins
for audio engineering on the Linux platform
[ TAP AutoPanner ] [ TAP Chorus/Flanger ] [ TAP DeEsser ] [ TAP Dynamics (Mono & Stereo) ] [ TAP Equalizer and TAP Equalizer/BW ] [ TAP Fractal Doubler ] [ TAP Pink/Fractal Noise ] [ TAP Pitch Shifter ] [ TAP Reflector ] [ TAP Reverberator ] [ TAP Rotary Speaker ] [ TAP Scaling Limiter ] [ TAP Sigmoid Booster ] [ TAP Stereo Echo ] [ TAP Tremolo ] [ TAP TubeWarmth ] [ TAP Vibrato ]
TAP DeEsser is a plugin for attenuating higher pitched frequencies in vocals such as those found in 'ess', 'shh' and 'chh' sounds. Almost any vocal recording will contain 'ess' sounds, whether a strong vocal delivery, from bad recording, speech impediments or simply many 'ess' words spoken together. Wind instruments and other musical instruments can also create shrill high-pitched noises. Audio engineers need to control these harsh 'ess' sounds in most recordings.
|I/O ports||1 input / 1 output|
|CPU usage (44.1 kHz)||5.9%|
|CPU usage (96 kHz)||12.8%|
|Hard RT Capable||No (see Notes)|
Audio input is fed into the audio path and the so-called Sidechain. In the Sidechain, the signal is filtered and then an attenuation value is computed based on the filter output. This computation is done using a hard-knee compressor characteristic with a threshold level set by the "Threshold" control and a compression ratio of 1:2. The compressor attack/release time is 10 ms for instant operation.
The filter applied to the Sidechain signal has two modes of operation: Highpass and Bandpass. Highpass mode is useful for attenuating several different 'ess' sounds. Bandpass mode is more suitable for attenuating a specific high frequency. The chosen filter's frequency is set by the "Frequency" control.
General advice about setting the "Frequency" control: The sounds attenuated by this plugin can be fit into four categories: male 'ess', male 'ssh', female 'ess' and female 'ssh' sounds. 'Shh' settings are quite applicable for 'ch', 'th', and hard consonants, such as 't', 'd', and 'k' as well. The following table lists frequencies for these four types of sounds. Note that these are only provided as a starting point for your own experimentations, as the characteristic frequencies of individual vocalists can vary in a suprisingly wide range. Use your ears!
|Male 'ess'||4500 Hz|
|Male 'ssh'||3400 Hz|
|Female 'ess'||6800 Hz|
|Female 'ssh'||5100 Hz|
Generally, female 'ess' and 'shh' sounds vary more in frequency than those of males. Due to this situation, you may find that using the Sidechain filter in the Highpass mode may be more responsive. Otherwise, with the Sidechain filter in Bandpass mode, only a narrow frequency area is being responded to. So if a singer has esses with varying frequency, try the Highpass mode of the Sidechain.
When choosing between Sidechain modes (e.g. Highpass or Bandpass) and setting the "Frequency" it may be very helpful to hear the Sidechain signal, which controls the compression of the through-passing audio: this is why the Monitor control is provided. While listening to the Sidechain signal, try to set the controls so that almost everything you want attenuated can be well heard, and even more importantly, almost everything you want to keep intact doesn't show up in the Sidechain sound.
Another aid in setting the plugin is the "Attenuation" output meter: this shows the momentary attenuation applied to the signal. It shouldn't show a significant attenuation all the time. Instead, it should indicate that attenuation springs into action only occassionally.
|name||min. value||default value||max. value|
|Threshold Level [dB]||-50||0||10|
Setting the "Threshold" too low will result in an unpleasant muffled sound. Also, if the singer starts sounding as if she had suddenly lost her teeth, take this as a sure sign that your "Threshold" setting is way too low. Use the ever popular Bypass button and watch out for side effects like this.
This plugin is best on vocal tracks, although it can be applied to mixed tracks as well. However, you should expect an increased amount of side effects (especially muffling and pumping of the sound) when attempting to de-ess the vocals of a mixed track.
Internal variables that depend on user settings are re-computed only when the corresponding user settings are changed. This keeps the plugin from being hard RT capable, although the effect of this occassional re-computation is not really noticeable when looking at CPU usage measures.
Because the Sidechain processing algorithm involves converting from
linear to decibel values for every sample, a custom, very fast lin2db
function was implemented via a log10 table, which is computed and
stored when the plugin library is loaded by the host. Without this
hack (that is, calling
log10f() for every sample), CPU
usage would be almost exactly twice this much.