Talking Bass: Formant Filters and Vowel Sounds in Synths

What is a talking bass?

A talking bass is a synth bass whose tone moves through vowel shapes, so it seems to say "wow", "yoi" or "yah". The effect comes from a formant filter rather than from a vocal sample.

Speech works the same way. A formant is a broad spectral peak created by a resonance of the vocal tract, and phoneticians have long demonstrated "artificial vowels" by passing a click train through a pair of band-pass filters^[1]. Swap the click train for a sawtooth bass and you have the basis of every talking bass patch.

This technique overlaps with the wobble bass, which moves a low-pass filter rhythmically. A formant filter moves several peaks at once, which is why it sounds like a mouth instead of a filter sweep.

How does a formant filter make a synth talk?

A formant filter boosts two or more narrow frequency bands at the positions where a human voice resonates for a given vowel. Move those bands together and the ear hears the vowel change.

The first two formants, F1 and F2, are usually enough to identify a vowel. F1 is higher for open vowels like "ah" and lower for closed ones like "ee" and "oo", while F2 is higher for front vowels like "ee" and lower for back vowels like "oo"^[1]. The table uses average male values from the same source.

Two practical rules follow. The source must be rich in harmonics, since a filter can only emphasize what is already there, so use saws, squares or distorted wavetables. And the bass note should sit well below F1, otherwise the formant peaks have too few harmonics to grab and the vowel blurs.

How to build a talking bass in Vital

Vital's Formant filter model uses band-pass filters tuned to vowel formant frequencies and comes in two styles, AOIE and AIUO. Its Formant X and Y controls move the formants, Spread changes their separation, Formant Transpose shifts them by semitones and Peak sets their emphasis^[2]. In the source code, the two styles are laid out as A, O, I, E and A, I, U, O^[3].

1. Pick a bright source
   Load a saw or a square on oscillator 1, add 2-4 unison voices with light detune, and play in the lowest two octaves of your bass range. Mono and legato on.
2. Add distortion first
   Put light drive or a waveshaper before the filter in the chain, or use the oscillator's own distortion, so the formant peaks have plenty of harmonics to work with.
3. Choose the Formant model
   Set filter 1 to Formant, style AIUO for the darkest "oo" position, or AOIE for more open vowels. Raise Peak until the vowels are clear but not whistling.
4. Map a macro to Formant X
   Assign macro 1 to Formant X and sweep it by hand. Find the two positions that sound most like words, for example "oo" and "ee".
5. Automate or modulate the macro
   Drive the macro with an LFO, an envelope or DAW automation (see the next section). This is where the bass starts talking.
6. Keep a clean sub underneath
   Route a sine on a separate oscillator directly to the output, or use a second bass track, so the formant movement does not change the low end.

How do you morph between vowels with an LFO or envelope?

Route an LFO to the formant position for repeating, rhythmic speech, and an envelope for a vowel move that happens once per note. Most talking basses use both: an envelope for the word and an LFO for the wobble inside it.

Vital's LFOs can run in Trigger mode, which restarts on each note, Sync mode, which locks to the host tempo, or Envelope mode, which plays the drawn shape once like a one-shot^[4]. Envelope mode turns any drawn LFO shape into a custom vowel contour. Kilohearts suggests the same trick for its Formant Filter: modulating either axis of the vowel pane with an LFO adds human-like movement^[5].

• "Wow" Envelope from "oo" to "ah" and back over one beat. Draw it as a rounded hump in an LFO set to Envelope mode.
• "Yoi" Envelope from "oo" or "oh" up to "ee" over 100-250 ms, usually with a small upward pitch bend at the same time.
• "Yah-yah" Synced LFO at 1/8 or 1/16 bouncing between "ee" and "ah". Retrigger on so every note starts on the same vowel.
• Stepped speech Sample-and-hold or a stepped drawn LFO at 1/16 jumping between three or four vowels. Sounds robotic and rhythmic.
• Performed Map the formant position to the mod wheel and record it. Hand-played vowel moves sound the most like real speech.

How do you make yoi and growl basses?

A yoi is a single vowel sweep per note, from a dark closed vowel to a bright one. A growl is a denser, distorted texture where the formant or vowel movement happens inside heavy FM and saturation. Both rely on the same formant filter, used at different speeds and levels of grit.

Yoi bass recipe

Start with a square or saw with moderate distortion. Set the formant filter to "oo" at rest, then route an envelope with fast attack, 150-250 ms decay and zero sustain so it sweeps to "ee" and back. Add a pitch envelope of +2 to +5 semitones on the same timing so the pitch rises with the vowel. Short notes on offbeats give the classic call-and-response yoi riff heard in riddim and midtempo.

Growl bass recipe

Growls start with harmonic chaos: FM between two oscillators, a wavetable with vocal-like frames, and distortion before the filter. Put the formant filter after the distortion, modulate its position with a synced LFO, then distort again and follow with multiband compression. The Serum bass design guide covers the FM and wavetable side of growls in detail.

Which synths and plugins have formant filters?

You can build a talking bass for free: Vital has a formant filter inside the synth, and Kilohearts gives away a formant filter plugin that works on any sound. Many paid synths also include vowel or formant filter modes.

How do you mix a talking bass?

Keep the sub separate and treat the talking layer as a midrange instrument. The vowel peaks live roughly between 250 Hz and 2.5 kHz, which is exactly where vocals, snares and leads compete.

High-pass the formant layer around 80-120 Hz, keep everything below that in mono, and use a dynamic EQ band on the loudest formant peak so bright vowels like "ee" do not jump out. If the track has a singer, duck the bass mids briefly under the vocal or push the formant movement into the gaps between lines. For a comparison with actual voice-driven effects, the vocoder vs talk box vs Auto-Tune guide explains where each one fits.