Praat

Praat
Name	Praat
Developer	Paul Boersma and David Weenink
Released	1991
Programming language	C, C++, Eiffel
Operating system	Microsoft Windows, macOS, Linux
License	Proprietary freeware (source available for some components)

Praat Praat is a software tool for phonetic analysis used in speech science, linguistics, and audiology. It provides waveform display, spectrogram analysis, formant tracking, pitch detection, and scripting for automated processing. The program has been cited in research across disciplines, used by investigators affiliated with institutions such as Max Planck Institute for Psycholinguistics, MIT, University of Edinburgh, Stanford University, and University of California, Berkeley.

Overview

Praat originated in the early 1990s and was developed by Paul Boersma and David Weenink at University of Amsterdam. It targets researchers working on topics such as phonetics, phonology, speech perception, speech production, and acoustics. The tool supports analysis techniques relevant to studies involving formant, fundamental frequency, spectral tilt, harmonics-to-noise ratio, and intensity measurements, and it is commonly used alongside laboratory equipment from vendors like Brüel & Kjær and National Instruments in projects funded by agencies such as the European Research Council and National Science Foundation.

Features and Functionality

Praat offers a suite of signal-processing functions including waveform viewing, spectrogram computation, and pitch-tracking algorithms rooted in autocorrelation and cepstral methods. Users can extract measures used in experiments on voice onset time and vowel space, compute formant contours with linear predictive coding, and perform time-alignment tasks related to forced alignment when paired with toolkits such as HTK or Kaldi. The scripting language allows batch processing for large corpora from archives like Linguistic Data Consortium and ELRA, and integrates with statistical workflows that utilize R (programming language), Python (programming language), MATLAB, and Praat TextGrid annotations interoperable with ELAN and Transcriber.

Architecture and Implementation

The core of Praat is implemented in compiled languages to deliver efficient digital signal processing on architectures supported by POSIX-compatible systems and mainstream desktop platforms. It uses internal data structures for objects such as Sound, Spectrum, Formant, Pitch, PointProcess, and TextGrid, enabling modular operations and plugin extensions. Script functions expose low-level controls for windowing, filtering, source-filter modeling, and inverse filtering, facilitating integration with toolchains used in projects at centers such as Harvard University, Columbia University, University of Oxford, and University of Toronto.

Applications and Use in Research

Researchers employ Praat in studies spanning sociophonetics, clinical phonetics, forensic phonology, and speech therapy. It is used to analyze accents and dialects from corpora like the BNC and Corpus of Contemporary American English, to quantify prosodic features in investigations at University of Cambridge and Yale University, and to measure pathological voice parameters in clinical trials at Mayo Clinic and Johns Hopkins Hospital. Praat-based methodologies appear in publications in journals such as Journal of the Acoustical Society of America, Phonetica, Language, and Cognition and feature in conference presentations at ICASSP, Interspeech, and ICphS.

Development and Licensing

Development has been driven by its original authors with contributions from a distributed user community; the project follows a release model with binary builds for Windows, macOS, and Linux and source code availability for inspection and limited redistribution. Licensing permits free use for academic and personal purposes while maintaining restrictions on commercial redistribution, and the software’s maintenance has been discussed in venues hosted by institutions like ESCA and societies such as the Acoustical Society of America.

Community and Resources

An active community of educators, researchers, and clinicians shares scripts, tutorials, and extensions via mailing lists, workshops at universities like University of Pennsylvania and University College London, and repositories hosted by research groups at Max Planck Society and regional consortia. Documentation includes the program’s built-in manual, user-contributed tutorials taught in courses at University of Oslo, University of California, Los Angeles, and training sessions at conferences like ACL and LREC. For reproducible research, practitioners combine Praat with data-management platforms such as GitHub, Zenodo, and institutional archives at DANS and UK Data Service.

Category:Speech processing software