Formants are usually associated with vowel sounds; they are the natural resonances of the vocal tract that give vowels their character. Acoustic instruments also have formants, which are derived from the body of the instrument. Many synths now come with formant filters and can be used in many ways to add character to sounds in the way a vocal tract or an instrument body does. I've used VirSyn TERA 3 here, but any synth or filter plug-in with parallel bandpass filters can do the job.
The three lowest-frequency vocal formants are the most important in producing intelligible vowel sounds. For that reason, formant filters on synthesizers are typically composed of three resonant bandpass filters working in parallel. The lowest three formants also characterize acoustic instruments. Vowel formant frequencies vary considerably, but the first three fall roughly into the 500, the 1,500, and the 2,500 Hz range, respectively (see Fig. 1). That's a good place to start when setting up your synth's formant filters. (For a detailed look at the role of formants in speech synthesis, see the article “Voices from the Machine” in the February 2004 issue of EM, available at www.emusician.com.)
FIG. 1: TERA 3''s formant filter is at the upper right of the control panel. The center display shows the three
formants for the vowel A.
Three's a Crowd
First, find a harmonically rich, bright, and sustained sound, such as TERA 3's PadSounds 04, that doesn't already use a formant filter. Next, turn off any active effects, and then modify the preset by inserting the formant filter just before the amplifier in the signal path. If there's another filter in the signal path, you might want to take it out.
Ensure that the formant filter's output is fully wet, and then set the low-, mid-, and high-band levels with boosts of 6, 3, and 0 dB, respectively — lower-pitched formants are usually louder. Formants are fairly broadband, so set the bandwidths, or resonance (as in TERA 3), to low values. A good way to temporarily focus on the effect of individual bands is to set the levels of the other bands to their minimum values (-50 dB in TERA 3).
A Nice Gesture
Vocal formants move around as you change the shape of your throat and mouth (that's how we speak), whereas instrument formants are mostly fixed. You can use fixed formants to alter the character of any sound, but synth filters are made for motion, and that's where things get interesting.
Moving one band of a formant filter relative to the others can have a profound effect. Set up an envelope generator with a short delay, a medium attack, and a full sustain, and then use it to modulate each formant's frequency up or down. Next, apply a single envelope, with opposite polarities, to the bottom and top formants. In TERA 3 you can use the envelopes by themselves or multiplied by Velocity, which results in more motion for louder notes.
Once you've enveloped the outer bands, apply an LFO to the middle band. Try multiplying the LFO by a delayed envelope to bring in that effect after the outer bands have settled down. Also try modulating the LFO rate with Aftertouch or the Modulation Wheel. In TERA 3 those are all source options in the modulation matrix.
A Step in Time
Modulating the band levels can also be effective. Multisegment envelopes or breakpoint automation with different rhythmic patterns for each band make good sources (see Web Clip 1). Bandwidth (resonance) is another good modulation destination for multisegment envelopes, but be sure to start with the minimum initial setting and a modest modulation amount to avoid damage to your speakers and ears.
While instrument formants are not pitch dependent, you can use key tracking to make one or more formants follow pitch. Set the formant frequencies to harmonic intervals of the pitch of an oscillator or sample, and assign full key tracking. Cutting and boosting the formant bands can be effective.
Bypassing effects and other filters helps you focus on the action of the formant filter. But once you have the formants under control, combining pre- and postfilter effects and lowpass and highpass filters in parallel with the formant filter is fair game. So leave realism behind, and give your tired presets a new voice.
Len Sasso is an associate editor of EM. Visit his all-new Web site at