Praat Scripts

A collection of Praat scripts for acoustic analysis and speech manipulation, developed primarily for research on infant-directed speech (IDS).

Author: Juan David Leongómez

---

Scripts

Script	Description
syllable-nuclei-with-review.praat	Detect syllable nuclei; compute speech and articulation rates with a manual review step
extract_formants_IDS.praat	Extract $f_0$ and formant summary statistics from a folder of IDS .wav files
IDS_manipulation.praat	Generate 8 factorial acoustic manipulations ($f_0$ mean × $f_0$ SD × formants) from a single recording

---

syllable-nuclei-with-review.praat

Detects syllable nuclei in a WAV file and calculates two speech rate metrics:

• Speech rate: syllables / total duration
• Articulation rate: syllables / phonation time (pauses excluded)

After automatic detection, the script opens the Praat editor so the user can manually review and correct both the syllable tier and the utterance/silence segmentation before results are written to a CSV file.

Features

• Sonority envelope: intensity is computed on a bandpass-filtered copy of the audio (default 100–6000 Hz, adjustable in the form), reducing false positives from low-frequency noise and fricative energy.
• Pitch-based utterance validation (VAD): each interval classified as utterance is validated against the Pitch object; intervals without sufficient vocal-fold vibration are reclassified as silent.
• Interactive WAV file selection via the system file chooser.
• Results are appended to a cumulative CSV in the same folder as the audio, so all recordings in a folder accumulate in a single output file.
• Compatible with Linux, Mac, and Windows.

Adapted from

• De Jong, N. & Wempe, T. (2009). Praat script to detect syllable nuclei and measure speech rate automatically. Behavior Research Methods, 41(2), 385–390. https://doi.org/10.3758/BRM.41.2.385
• Quené, H., Persoon, I. & De Jong, N. (2010). Modified version of the original script [version 2010.09.17]. https://github.com/FieldDB/Praat-Scripts/blob/main/praat-script-syllable-nuclei-v2dir.praat

---

extract_formants_IDS.praat

Extracts $f_0$ (mean and SD) and formant frequencies (F1–F4) summary statistics from a folder of IDS speech .wav files, and writes one row per file to a CSV.

Parameters

Parameter	Description	Default
directory	Folder containing the .wav files	—
resultsfile	Full path of the output CSV	—
minimum_pitch / maximum_pitch	$f_0$ analysis range	100 / 600 Hz (female)
maximum_formant	Formant analysis ceiling	5500 Hz (female/children)

Notes

• Default settings are tuned for female speakers. For male speakers, use minimum_pitch = 75, maximum_pitch = 300, and maximum_formant = 5000.
• Undefined values (unvoiced or silent segments) are replaced with 0.
• Formant analysis uses the Burg method (5 formants, 25 ms window, pre-emphasis from 50 Hz), matching the settings of Hilton et al. (2022).

Adapted from

• Adapted from analysis/acoustics_processing/3_masterscript.praat in Hilton, C. B., Moser, C. J., et al. (2022). Acoustic regularities in infant-directed speech and song across cultures. Science, 378(6617). https://doi.org/10.1126/science.abm1720
• Original script repository: https://github.com/themusiclab/infant-speech-song

---

IDS_manipulation.praat

Generates all 8 combinations of a 2 ($f_0$ mean: High/Low) × 2 ($f_0$ SD: High/Low) × 2 (formant frequencies: High/Low) factorial design from a single IDS recording. Each manipulated file is saved as a .wav, and a verification report with pre- and post-manipulation acoustic values is saved as a .txt file.

Manipulation magnitudes are set to 1 × the between-speaker SD of IDS acoustics in female speakers reported by Hilton et al. (2022, Nature Human Behaviour; n = 21 societies):

Parameter	Offset	Default
$f_0$ mean	± n Hz	±51 Hz (50.8 Hz)
$f_0$ SD	± n Hz	±28 Hz (28.4 Hz)
Formants (F1–F4)	± proportion	±4.6% (CV = 0.046)

Manipulation procedure

1. Formant shift (Step 1): LPC-based resynthesis applied to the original audio. All formant frequencies are scaled proportionally by a ratio that is adjusted iteratively until the geometric mean of F1–F4 in the output matches the target within ffreq_tolerance_hz. No PSOLA is used in this step; pitch is left untouched.

2. $f_0$ manipulation (Step 2): Iterative affine correction via PSOLA. A linear transform (slope and intercept) is computed to shift $f_0$ mean and SD to their targets. After each resynthesis, the output SD is measured; if it deviates from the target by more than sd_tolerance_hz, the slope is corrected by target_sd / measured_sd and the process repeats. Typically converges in 2–3 iterations (maximum max_iter).

Parameters

Parameter	Description	Default
Input_file	Full path to the input .wav file	—
Output_dir	Output folder (trailing / required)	—
sd_f0mean_hz	$f_0$ mean offset in Hz	51
sd_f0sd_hz	$f_0$ SD offset in Hz	28
factor_formant	Formant scaling factor	0.046
pitch_floor_hz / pitch_ceiling_hz	$f_0$ analysis range	100 / 500 Hz
sd_tolerance_hz	Convergence tolerance for $f_0$ SD	0.5 Hz
max_iter	Maximum iterations for $f_0$	5
ffreq_tolerance_hz	Convergence tolerance for formants	10 Hz
max_ffreq_iter	Maximum iterations for formants	12
max_formant_synth_hz	Formant ceiling for synthesis and measurement	5500 Hz

Output files

• 8 .wav files named {name}_f0mean-{High/Low}_f0SD-{High/Low}_Ffreq-{High/Low}.wav
• 1 .txt report named {name}_mean±{n}Hz_sd±{n}Hz_formant±{n}pct.txt

Sources

• Manipulation magnitudes: Hilton, C. B., et al. (2022). Acoustic regularities in infant-directed speech and song across cultures. Nature Human Behaviour, 6, 1545–1556. https://doi.org/10.1038/s41562-022-01410-x
• Formant manipulation adapted from changeformants.praat, Praat Vocal Toolkit

---

Licenses

Each script carries its own license header because they derive from different upstream sources:

Script	License	Upstream
syllable-nuclei-with-review.praat	GPL-3.0-or-later	de Jong & Wempe (2008); Quené et al. (2010) via FieldDB/Praat-Scripts
extract_formants_IDS.praat	CC BY-NC-SA 4.0	Hilton, Moser et al. (2022) via themusiclab/infant-speech-song
IDS_manipulation.praat	GPL-2.0-or-later	Original code; formant shift procedure adapted from Praat Vocal Toolkit changeformants.praat (verify license)

Note: extract_formants_IDS.praat is released under CC BY-NC-SA 4.0, which restricts commercial use. The other two scripts are GPL and permit commercial use. These licenses are not mutually compatible, so a single repo-wide license is not applied.