Vowels

A morphing formant filter that moves through five vowel positions using selectable voice profiles and three filter-bank topologies.

What it does #

Vowels models speech-like resonances by tuning several filters to stored formant frequencies, gains, and bandwidths. Blend moves between as many as five chosen vowel positions, Profile selects the underlying vocal data, and Structure changes how the formant filters are connected.

Controls and ports #

Control	Label	Official description
blend	Blend	Blends the surrounding vowels '-100 %', '0.00 %', and '100 %' will always sound a single vowel position (1, 3, or 5). At '-50.0 %' or '50.0 %', positions 2 or 4 will be heard (or when set to "None", a blend of the neighboring vowels).
vowel	Vowel	Vowel sound set for this position:
profile	Profile	Vowel data set:
reso	Reso	Amount of feedback applied within the filters
pitch	Pitch	Pitch-based (musical) detune of the filter bank
freq	Freq	Linear (Hertz) detune of the filter bank
structure	Structure	Filter circuit:
vowel_in	Vowel In	Control input added to Vowel Blend

Practical uses #

• Feed it a harmonically rich saw, pulse, noise, or vocal-like source and sweep Blend for articulated movement.
• Assign contrasting vowels to the five positions, then modulate Blend slowly or from note expression.
• Use Pitch for proportional formant movement and Frequency for a fixed shift when designing robot or creature voices.

Things to know #

• Blend -100, 0, and +100 percent land on positions 1, 3, and 5; positions 2 and 4 occur at -50 and +50 percent when assigned.
• Pitch multiplies all formant frequencies, while Frequency adds Hertz before the final proportional pitch scaling; they are not interchangeable.
• Cascade, Parallel, and Parallel with high-pass contribution produce substantially different level and spectral behavior.
• The implementation smooths vowel data and topology changes, so fast modulation does not necessarily follow sample by sample.

Technical details #

A high-confidence Nitro implementation candidate supports the technical summary, but the visible-module mapping remains inferred.

The matching filter/Vowel body reads four primary formants per profile/vowel from an internal data table and interpolates them piecewise across five positions. At Blend 0 it selects position 3; negative and positive halves interpolate through positions 2/1 and 4/5, skipping an unassigned intermediate position when necessary. It smooths frequency, gain, Q, topology, and formant distance at different rates, then runs at 4x oversampling. Cascade uses serial low-pass SVF sections and extends the upper response with additional stages up to 12; Parallel sums four formant filters, with its alternate topology adding a high-pass contribution. Resonance scales Q, Shift adds Hertz, and pitch/frequency scaling multiplies the formant bank.

Nitro port	Type
in	InputAudioPortType
freq	InputAudioPortType
resonance	InputAudioPortType
blend	InputAudioPortType
freqMod	InputAudioPortType
freqModAmount	InputAudioPortType
drive	InputAudioPortType
profile	InputValuePortType
vowel	InputValuePortType
invClipLevel	InputAudioPortType
shift	InputAudioPortType
topology	InputValuePortType
out	OutputAudioPortType

Open questions #

• Document every Profile, vowel symbol, and Structure label and verify their ordering against the stored data tables.
• Measure formant frequencies, interpolation, Q/gain normalization, topology crossfades, and modulation smoothing.
• Test stereo Blend offsets, Drive limiting, high resonance, extreme Pitch/Frequency shifts, latency, and CPU scaling.

Version and sources #

Checked against Bitwig Studio 6.0.6. This page combines Bitwig's module metadata, the existing Grid course and guides, and Nitro analysis where the mapping is strong enough to support a technical statement.

Related material #

• Grid Modules course reference
• Filter+ - Modular Polyphonic Distortion & Filter Effect

Return to Bitwig Grid Modules or the Grid Modules course lesson.