Praat Tutorial and Resources
Jean-Philippe Goldman
February 2004
The purpose of this document is to give an overview of Praat, its features, its on-line resources and some quick steps to start using it.
What is Praat?
Praat is an open-software tool for speech signal edition and labeling, as well as for various acoustic analyses (spectral, formant, pitch, intensity) and prosodic manipulation. It has also some abilities to plot figures, build Optimality-Theory grammars, make articulatory synthesis, simulate feed-forward neural networks and make discriminative statistics. Paul Boersma and David Weenink from the Institute of Phonetic Sciences, University of Amsterdam (www.fon.hum.uva.nl), have been developing Praat for almost 10 years and are still very active (frequent software updates and quick answers on praat-users group). It's free and available for most platforms. Many tutorials exist and the discussion group is very active (see resources below).
It has been designed for both beginners (easy interface, many default options to learn by trying, searchable manual, various possibilities of analysis, manipulation and labeling) and expert users (logging of every action for easy batch scripting, or resource for advanced application development\).
Quick Tutorial
Praat's layout is based on 'objects'. Any object (sound, pitch curve, matrix, group of strings,...) can be loaded, created, saved, queried, modified or used to create another object (of the same type or not).
General information about this tutorial:
This tutorial will teach you how to create or load sounds, to make first analyses and to label them. Praat needs no right-click button nor double-click and has no contextual menus. Follow the instructions in red.
Figure 1 Praat icon
Start Praat
Praat starts with two windows: Praat objects, and Praat picture (you can close the second one as we will not draw pictures here). The "Praat objects" window is the main one. It lists the objects in memory. Menus and buttons are dynamic, i.e. they may change (appear, disappear or be disabled) according to the selected objects. When it's just started, Praat has no objects in its list, therefore the list is empty, the buttons
(bottom) are yet disabled, and the 'Write' menu (4th) is empty.
Figure 2 Praat windows at start
Create a Sound
Execute Menu 'New' > Sound > Create Sound from > OK
The sound you’ve just created is now in the list of objects (type + name). As it is selected (grey), many buttons appear on the right, the buttons at bottom are now enabled, and the Write menu displays options.
In fact the 5 buttons at the bottom are called the fixed buttons (they are always present though they might be disabled if nothing is selected or present in the list). These functions are common too all objects whatever the type. With them, you may:
Rename an object (Rename…)
Duplicate an object (Copy…)
Delete an object (Remove)
Get the results from various queries on an object (Info)
Browse the internal data of an object (Inspect)
The buttons on the right of the list are the dynamic buttons. We'll go through most of them. The first one is always a Help button referring to the type of the selected object (it won't appear if several objects of different types are selected at the same time).
Figure 3 "Praat objects" window with "Sound" objects
Record a Sound
With a microphone, you can record a sound or some speech with Menu > New > Record A window 'SoundRecorder' opens (see below). Plug your microphone. Press Record to start recording. You should see a real time meter in green/yellow/red while you speak. If not, check your general recording settings. Press Stop to stop recording. Play to listen to your recording. You can start over if you are not satisfied. When finished, press Save to list: (change the name 'sound' to something else if you like) then Close. The sound appears now in the object list.
Figure 4 SoundRecorder window
Load a Sound
From Read menu> Read , Praat reads whatever it can recognize. For sound files, WAV, AIFF, non-compressed AIFC, NeXT/Sun (.au) and NIST files are supported. See /manual/Sound_files_3__Files_that_Praat_can_read.html
Raw (headerless) files can also be read with other options in this Read menu.
Sound Object
Once you have a Sound object in your list (created, recorded or loaded), select it by clicking on it if it's not selected (see figure 3). Here is a quick description of the dynamic buttons:
Sound Help opens the Help window with many useful information
Edit opens an Editor window with signal and various acoustic analyses. We’ll describe it in the next section
Play plays the sound. Press the Esc key to stop playing.
object toDraw- opens a local menu with options. It is useful to plot picture in the window we've closed. You can give a try.
Query- opens a local menu with many options. Try some of them, like An Info window will open to display results. Try also Get nearest A form window asks for a time value. In this form, notice the Help button referring to the right page in the Manual. Try Info fixed button (at bottom); the info window will give you all the possible queries on the whole file.
Modify- allows to Reverse the sound file (try it and Play). Notice that modifications are done with the object. Reverse it once more to get back to original sound.
Modify- > Formula. See Formulas tutorial for more help (from Help menu, or Help button once you or see /manual/Formulas.html).
Annotate- > To TextGrid… create a labeling object (see TextGrid object section)
The Analyze section or submenu is a group of dynamic buttons that create new objects of different types from the sound object after one acoustic analysis among various explicit possibilities:
Periodicity- deals with pitch contour: Pitch analysis, Harmonicity, Glottal pulses Spectrum- deals with
all kind of spectral analyses (spectrum, spectrogram, wavelets) Formant&LPC- creates formant contours and (linear predictive) filter coefficients
To Intensity calculates to intensity contour
When you try these tools, notice that the dynamic buttons change. Click the Edit dynamic button to see results. Reselect the original Sound object to try another one. The Manipulation section contains a unique button:
To Manipulation:creates a Manipulation object, which contains the original sound, a pitch tier, a duration tier and a PointProcess (i.e. a sequence of points in time) representing the glottal pulse.
Most of the buttons in the Synthesize section are self-explicit and create Sound objects from the original one(s). You'll find here tools to extract a part, to resample, to filter (see Filter tutorial in Help Menu), or to concatenate sounds.
Edit Window
As a Sound object is selected, press Edit, a new window opens. In the top half, see the speech signal in black as well as the glottal pulses. The bottom half shows the analyses: pitch in blue, intensi
ty in yellow (or green), formants in red and spectrogram in background. Each of the five analyses (these four and the pulses) has its own menu, all starting with a ‘Show/Hide’ switch and a ‘Settings…’ option.
The three bars below the analysis pane are ‘play’ buttons. You can click them to play the sound and see a real time cursor (vertical bar moving from left to right) on the signal. The top cursor bar is divided in two equal sections as the main cursor or marker is set by default right in the middle of the file when Edit window opens. Figures indicate durations in seconds. The second bar (window bar) represents the part of the file that is shown in the window. The third is the entire file (full bar). The last two are now the same as Edit window shows all the file when it opens.
Select a part in the middle of the signal by a click-and-drag on signal or analysis pane. The selection appears in transparent pink. Now the top bar should be divided in three sections: before the selection, the selection itself, after the selection. Play them individually. You can expand or shrink this selection by pressing-and-holding the shift button before you click the signal pane or make another selection.
Now look at the bottom left corner, you see four handy buttons to help you modify the window settings
all shows the entire file
in let you zoom in by a factor of two (i.e. you’ll see half of the current window)
out zooms out by a factor of two
sel adjusts the display window to the selection
Try them. As the window displays a portion of the signal, notice how the window bar and the full bar change. You can still play what is before the window, after the window and the complete sound.
Figure 5 Sound Edit window
As you click on the signal, see in blue, on the left of the sound, the value of the waveform. If you click on the analysis pane, you may have up to three vertical scales depending on the number of shown analyses:
frequency on the left. It shows the value in Hz of the marker in the frequency domain (as long as spectrum or formant analysis is shown)
pitch in blue on the right. marker value as well as minimum and maximum. Note that the marker value is the pitch estimation of the sound at that moment, i.e. on the blue curve, not at the marker position which can be anywhere in the pane.
intensity in green. It might be either side, or even inside the analysis pane.
Here is briefly what you can do with the other menus:
File: extracts parts of the signal to the object list or to a file
Edit: the usual Undo/Cut/Copy/Paste menu for signal edition. Reverses and sets to silence options too.

版权声明：本站内容均来自互联网，仅供演示用，请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系QQ:729038198，我们将在24小时内删除。