Vowel Formant Filter (a-e-i-o) — User Guide

Simple vowel formant filtering: applies four vowel formant characteristics (a, e, i, o) to any sound using LPC filtering, then concatenates the results — creates vocal-like timbral transformations.

Author: Shai Cohen Affiliation: Department of Music, Bar-Ilan University, Israel Version: 0.1 (2025) License: MIT License Repo: https://github.com/ShaiCohen-ops/Praat-plugin_AudioTools

Contents:

What this does Quick start Vowel Characteristics Applications

What this does

This script applies vowel formant filtering to transform any sound through four vowel timbres. Process: (1) Creates vocal tract models for vowels "a", "e", "i", "o" using Praat's built-in phone models. (2) Converts each vocal tract to LPC (Linear Predictive Coding) filter. (3) Filters the input sound through each vowel's LPC filter separately. (4) Concatenates all four filtered versions (a→e→i→o sequence). Result: Four variations of the input sound, each colored by a different vowel's formant structure. No parameters — fully automatic. Output duration is 4× original duration (four consecutive vowel-filtered versions).

What are formants? Formants are resonant frequencies of the vocal tract that define vowel identity. Each vowel has a characteristic pattern of formant frequencies (F1, F2, F3). For example: "a" (as in "father") has low F1 (~700Hz), low F2 (~1200Hz) — open, back vowel. "i" (as in "see") has low F1 (~300Hz), high F2 (~2300Hz) — closed, front vowel. This script applies these formant characteristics to any sound via LPC filtering, creating "vowel-colored" timbres. Common in vocoders, talkbox effects, and experimental sound design.

Quick start

In Praat, select exactly one Sound object.
Run script… → a-e-i-o_filter.praat.
No parameters — script runs automatically.
Output: "VT_concatenated_[samplerate]" (4 filtered versions concatenated).
Individual filtered sounds also available: "VT_a_[sr]", "VT_e_[sr]", "VT_i_[sr]", "VT_o_[sr]".
Result auto-plays after processing.

Quick tip: Works on any sound — synthesizers, drums, noise, speech. For speech, creates "talking robot" effect. For percussive material, adds vowel-like resonances (e.g., kick drum becomes "boom-beem-boop-bohm"). Output is 4× longer than input. To use individual vowels, select them separately from Objects window (don't use concatenated version). Processing instant (<1 second for typical audio).

Important: No user parameters — vowels and order are fixed (a→e→i→o). For custom vowel selection or different order, modify script directly. LPC filtering can alter amplitude significantly — output may be louder or quieter than input. Check levels and normalize if needed. Script auto-cleans intermediate objects, keeping only original + outputs.

Vowel Formant Characteristics

The Four Vowels

Vowel "a" (as in "father")

Formants: F1 ≈ 700Hz, F2 ≈ 1200Hz, F3 ≈ 2500Hz

Character: Open, back vowel. Dark, warm timbre. Low-mid frequency emphasis.

Effect on sound: Adds warmth, fullness. Emphasizes fundamental and lower harmonics. Creates "aaah" quality.

Vowel "e" (as in "bed")

Formants: F1 ≈ 500Hz, F2 ≈ 1800Hz, F3 ≈ 2500Hz

Character: Mid-front vowel. Balanced, present timbre. Mid-frequency emphasis.

Effect on sound: Adds clarity, presence. Emphasizes mid-range. Creates "eh" quality.

Vowel "i" (as in "see")

Formants: F1 ≈ 300Hz, F2 ≈ 2300Hz, F3 ≈ 3000Hz

Character: Closed, front vowel. Bright, thin timbre. High-frequency emphasis.

Effect on sound: Adds brightness, edge. Emphasizes upper harmonics. Creates "eeee" quality.

Vowel "o" (as in "go")

Formants: F1 ≈ 400Hz, F2 ≈ 800Hz, F3 ≈ 2600Hz

Character: Mid-back vowel. Rounded, mellow timbre. Low-mid resonance.

Effect on sound: Adds roundness, mellowness. Darker than "a" but not as open. Creates "ohhh" quality.

LPC Filtering Process

Technical implementation:

Vocal Tract models: Praat's built-in phone models define formant patterns for each vowel
VocalTractTier: Creates time-varying vocal tract (static in this case — 0.5 value throughout duration)
LPC conversion: 5ms analysis window converts vocal tract to Linear Predictive Coding filter
Filtering: Input sound convolved with LPC filter (no pre-emphasis)
Resampling: Ensures output matches original sample rate (quality 50 = high)

Applications & Use Cases

Vocal-like Effects

Material: Synthesizers, pads, drones

Result: Adds vowel-like resonances to electronic sounds. Creates "singing synthesizer" or "vocal formant" effect without actual voice. Useful for: Talkbox-style sounds, vocoder-like textures, humanizing electronic timbres.

Percussive Transformation

Material: Drums, percussion, impacts

Result: Each hit acquires vowel character (kick drum → "boom", "beem", "boop", "bohm"). Creates pitched/tonal quality from transients. Useful for: Experimental drum processing, "talking drums" effect, timbral variation in rhythm.

Vowel Morphing

Technique: Concatenated output creates a→e→i→o progression

Result: Smooth timbral journey through vowel space. Useful for: Sound design, transitions, demonstrating formant changes, creating evolving textures.

Speech Robotization

Material: Voice recordings

Result: Re-filters voice through standardized formants, removing natural variation. Creates "synthetic voice" or "vocoder" character. Useful for: Robotic dialogue, experimental voice processing, sound art.

Comparison to Related Effects

Effect	Method	vs Vowel Formant Filter
Vocoder	Carrier + modulator with band analysis	Vowel filter: Simpler, fixed formants, no carrier needed
Talkbox	Physical vocal tract filtering via tube	Vowel filter: Digital simulation of talkbox concept
Formant filter	Parametric bandpass filters at formant frequencies	Vowel filter: Uses LPC (more accurate vocal tract model)
EQ	Static frequency response shaping	Vowel filter: Dynamic formant peaks, vocal-specific resonances

Creative Variations

Using individual vowel outputs:

Select only "VT_i_[sr]" for bright, aggressive character
Select only "VT_a_[sr]" for warm, full-bodied timbre
Layer multiple vowels for complex formant interactions
Process individual vowels further (reverb, delay, distortion)

Script modification ideas:

Change vowel order: Edit lines 79-82 to reorder concatenation (e.g., i→e→a→o)
Add more vowels: Create additional vocal tracts ("u", "æ", "ə") and filter/concatenate
Adjust LPC window: Change 0.005 (line 47) to alter formant tracking (0.001-0.01 range)
Skip concatenation: Comment out lines 79-84 to keep only individual vowel outputs

Troubleshooting

Problem: Output too quiet or loud
Cause: LPC filtering alters amplitude based on formant energy
Solution: Normalize output (Scale peak to 0.99 in Praat) or adjust input level before processing

Problem: Output sounds distorted or clipped
Cause: Formant resonances amplify certain frequencies beyond 0dB
Solution: Reduce input amplitude to 0.5-0.7 before processing, or normalize after

Problem: Vowel effect too subtle on some material
Cause: Input lacks harmonic content (pure noise, sine waves)
Solution: Works best on harmonically-rich material (sawtooth, square waves, complex tones). For noise, effect creates subtle coloration only.