liminfoPraat Reference
Free reference guide: Praat Reference
About Praat Reference
The Praat Script Reference is a searchable guide to Praat, the leading software for phonetic and speech analysis. It covers spectrogram display settings and Sound-to-Spectrogram conversion, formant tracking using the Burg algorithm with gender-appropriate frequency ceilings (5000 Hz male, 5500 Hz female/child), and formant value queries at specific time points.
Pitch analysis sections include F0 extraction via autocorrelation and cross-correlation methods, mean and standard deviation queries. Voice quality measurements cover jitter (period perturbation, normal < 1.040%), shimmer (amplitude perturbation, normal < 3.810%), and HNR (harmonics-to-noise ratio, normal > 20 dB) using PointProcess glottal pulse detection.
The reference also documents TextGrid annotation workflows (boundary insertion, interval labeling, multi-tier creation), file I/O operations (WAV/AIFF/MP3/FLAC reading and writing), audio editing (Extract part, bandpass filtering, resampling), and Praat scripting constructs including for loops, if/endif conditionals, batch file processing, and CSV result export with appendFileLine.
Key Features
- Spectrogram settings for wideband (0.005s) and narrowband (0.03s) analysis with Sound-to-Spectrogram conversion
- Formant tracking with Burg algorithm including male (5000 Hz) and female/child (5500 Hz) ceiling parameters
- F0 pitch extraction via autocorrelation and cross-correlation with mean and standard deviation queries
- Voice quality metrics: jitter (< 1.040%), shimmer (< 3.810%), and HNR (> 20 dB) with normal thresholds
- TextGrid annotation commands for boundary insertion, interval labeling, and multi-tier creation
- File I/O for WAV, AIFF, MP3, FLAC reading and TextGrid text file saving
- Praat scripting with for loops, if/endif conditionals, batch file processing, and appendFileLine CSV export
- Audio processing commands including Extract part, bandpass Hann filtering, and sampling rate conversion
Frequently Asked Questions
How do I extract formant values (F1, F2) in Praat?
Select your Sound object and run "To Formant (burg): 0.0, 5, 5500, 0.025, 50" where 5500 Hz is the maximum frequency ceiling (use 5000 Hz for male speakers). Then query values with "Get value at time: 1, 0.5, hertz, Linear" for F1 at 0.5 seconds. Use "List" to output F1/F2/F3 values as a table for the entire duration.
What is the difference between wideband and narrowband spectrograms?
Wideband spectrograms use a short window length (0.005s) providing good time resolution to show individual glottal pulses and formant bands. Narrowband spectrograms use a longer window (0.03s) providing better frequency resolution to reveal individual harmonics. Set these via "View & Edit > Spectrogram settings" or the "To Spectrogram" command.
How do I measure F0 (fundamental frequency) in Praat?
Use "To Pitch: 0.0, 75, 600" for autocorrelation-based pitch extraction, where 75 Hz is the floor and 600 Hz is the ceiling. For cross-correlation, use "To Pitch (cc)". Query the mean F0 with "Get mean: 0, 0, hertz" and standard deviation with "Get standard deviation: 0, 0, hertz". Adjust floor/ceiling based on speaker gender.
How do I measure jitter, shimmer, and HNR for voice quality assessment?
First create a PointProcess from Sound + Pitch with "To PointProcess (cc)". Measure jitter with "Get jitter (local): 0, 0, 0.0001, 0.02, 1.3" (normal < 1.040%). For shimmer, select Sound + PointProcess and use "Get shimmer (local)" (normal < 3.810%). For HNR, use "To Harmonicity (cc): 0.01, 75, 0.1, 1.0" then "Get mean" (normal > 20 dB).
How do I create and edit TextGrid annotations?
Select your Sound and run "To TextGrid: sentence word phone, empty string" to create three annotation tiers. Insert boundaries with "Insert boundary: 1, 0.5" (tier 1 at 0.5 seconds). Label intervals with "Set interval text: 1, 2, hello" (tier 1, interval 2). Save with "Save as text file" to preserve annotations in Praat TextGrid format.
How do I write a Praat script to batch process multiple WAV files?
Use Create Strings as file list to get all WAV files, then loop: "for i from 1 to n / selectObject files$ / file$ = Get string: i / Read from file: file$ / endfor". Within the loop, perform your analysis and export results with "appendFileLine: results.csv, file$, comma, f0_mean, comma, f1_mean" to build a CSV output file.
How do I apply a bandpass filter or resample audio in Praat?
For bandpass filtering, select your Sound and use "Filter (pass Hann band): 300, 3000, 100" to keep the 300-3000 Hz range with 100 Hz smoothing. For resampling, use "Resample: 16000, 50" to convert to 16 kHz with precision 50. Extract a time segment with "Extract part: 0.5, 1.5, rectangular, 1, no" for the 0.5 to 1.5 second interval.
What audio file formats does Praat support?
Praat supports reading WAV, AIFF, MP3, and FLAC audio files with "Read from file: path/to/file.wav". For saving, use "Save as WAV file" for audio output. TextGrid annotations are saved with "Save as text file" in Praat native format. The scripting engine supports batch processing across hundreds of files using file list strings and for loops.