praat / praat Goto Github PK

Hi, as I stated in the title, I can't find the xcode project file or any zip file containing it like you said in the compiling guide.
So please guide me what I should do if I want to compile Praat in mac os. Sorry for my bad English anyway.

The interval aligner does not currently work in Linux (Kubuntu 15.10, 32-bit). In 6.0.15 it creates empty intervals, and in 6.0.16 it crashes Praat with a message about non-matching sampling rates.

See the thread on the mailing list for more details.

As requested, @PaulBoersma

Currently (6.0.15), inline comments (marked by ;) in lines with command names are interpreted to be part of the command name:

Create TextGrid: 0, 1, "tier", "" ; Good
asserterror Command "Remove ; Bad!"
  ... not available for current selection.
Remove ; Bad!

As shown in the snippet, a comment next to the first command (after the argument list) is correctly ignored, but when used in a line with a command name without arguments, it is taken to be part of the command name, and raises an error.

Is this the correct behaviour?

Praat 5.4.22 and 6.0.03 generate different text versions of FFNet objects, and at least Praat 6.0.03 complains when opening an FFNet object saved by 5.4.22.

The old objects looked like this:

File type = "ooTextFile"
Object class = "FFNet"

nLayers = 1 
nUnitsInLayer []: 
    nUnitsInLayer [0] = 4 
    nUnitsInLayer [1] = 3 
outputsAreLinear = 0 
nonLinearityType = 1 
costFunctionType = 1 
outputCategories: size = 0 
nWeights = 15 
w []: 
    w [1] = -0.07111157301884496 
    w [2] = -0.007799936451578726 
    w [3] = -0.0736766578690795 
    w [4] = -0.05769455299999926 
    w [5] = -0.0038267067826605497 
    w [6] = 0.056617462520272915 
    w [7] = -0.06044512424540989 
    w [8] = -0.04086647532466166 
    w [9] = -0.030626591089634614 
    w [10] = 0.01798069280191066 
    w [11] = 0.03951275934685821 
    w [12] = 0.06319214749062219 
    w [13] = 0.06850375664532089 
    w [14] = -0.052898212612844664 
    w [15] = 0.04948936042643251 

The objects saved by Praat 6.0.03 replace the line that says outputCategories with this:

outputCategories? <absent> 

and Praat complains saying

Found a number while looking for an enumerated value in text (line 11).
FFNet not read.

This problem has been reported as a bug report on the Debian package for praat (https://bugs.debian.org/859343). I am forwarding it here, hoping that you know how to fix it.

Best,

Rafael Laboissière

Running Praat from the command line through prove results in Praat incorrectly trying to open the test scripts instead of running them.

prove is a Perl-based utility to run TAP tests. As such, it makes it possible to use any interpreter through its --exec option, and not just Perl itself. However, this does not work in Praat v6.0 at least.

Here's a test case:

# cpanm Test::Harness or sudo apt-get install libtest-harness-perl
# but prove should come with the standard Perl distribution
git clone http://gitlab.com/cpran/plugin_testsimple.git
cd plugin_testsimple
prove --exec "praat"

The last command runs the tests (which by default match ./t/*.t) through praat, and should result in the output of each of the scripts being parsed by prove for general statistics. Instead, each test script is read and opened in the Praat editor in turn.

The problem is not related to the extensions of the individual test scripts (in this case, t). It is probably related to the way in which prove calls Praat internally.

I guess this should be fixed by improving the detection of the environment under which Praat is called, but in the meantime it could be fixed by implementing an --exec or --run option in Praat itself (as once planned) to force execution mode (bypassing the detection altogether).

BTW: This was using Praat 6.0 under Kubuntu 14.04, with Perl 5.22 and TAP::Harness 3.35

I'm trying to get praat accepted into a linux distribution but the licensing terms are unclear.

main/GNU_General_Public_License.txt is version 2.0, but under external there are several files with a gpl 3.0+ header.

Since I couldn't find anywhere a notice about the former being gpl 2.0 "or any later version", there are distributability issues with it being linked to the latter.

So, could you please clarify the question?

I would like to build Praat on OS X, but I'm falling at the first hurdle.
The instructions in README.md start with

Extract the *xcodeproj64.zip* file from the latest release

but where do I find this xcodeproj64.zip file? It's not in the praat-master.zip file that I downloaded, and I don't see it when I do a git clone either.

I tried to build the program using make and I got this.
I'm using Ubuntu 16.04

praat_Sound_init.cpp: In function ‘void INFO_Praat_reportSoundServerProperties(UiForm, int, Stackel, const char32_, Interpreter, const char32_, bool, void_)’:
praat_Sound_init.cpp:2156:1: error: ‘END’ was not declared in this scope
END
^
In file included from praat_Sound_init.cpp:19:0:
../sys/praat.h:484:157: error: a function-definition is not allowed here before ‘{’ token
rm sendingForm, int narg, Stackel args, const char32 *sendingString, Interpreter, const char32 *invokingButtonTitle, bool, void *okClosure) {
^
praat_Sound_init.cpp:2159:1: note: in expansion of macro ‘FORM_WRITE3’
FORM_WRITE3 (SAVE_Sound_saveAsAifcFile, U"Save as AIFC file", nullptr, U"aifc") {
^
praat_Sound_init.cpp:2770:1: error: expected ‘}’ at end of input
}
^
praat_Sound_init.cpp:2770:1: error: expected ‘catch’ at end of input
praat_Sound_init.cpp:2770:1: error: expected ‘(’ at end of input
praat_Sound_init.cpp:2770:1: error: expected type-specifier at end of input
praat_Sound_init.cpp:2770:1: error: expected ‘)’ at end of input
praat_Sound_init.cpp:2770:1: error: expected ‘{’ at end of input
praat_Sound_init.cpp:2770:1: error: expected ‘}’ at end of input
praat_Sound_init.cpp:2770:1: error: expected ‘}’ at end of input
: recipe for target 'praat_Sound_init.o' failed
make[1]: *_* [praat_Sound_init.o] Error 1
make[1]: Leaving directory 'cppworkspace/praat/fon'
makefile:14: recipe for target 'all' failed
make: *** [all] Error 2

a0b9da5 introduced a new default executable name for Praat when compiled as barren.

What is the rationale behind this? This change means that computers that currently use automated build systems using a barren praat (including the ones I maintain), will either have to be modified to change the name of the executable, or will have to use a version of scripts that is specifically written to use that executable.

As an aside, is there any reason why a change like this, which will most likely result in breakage, is introduced without any warning or notification whatsoever?

On Arch Linux, I would like to install praat. There is an aur PKGBUILD for doing this (actually three, https://aur.archlinux.org/packages/praat-pulse-git/ is the one that looks like I want it, but this issue applies to all three), but the compilation fails in praat_DataModeler_init.cpp – that file assumes the INFO_ONE macro is defined, but praat.h has no such #define.

On the filtering page there is a broken link to the page on band filtering in the frequency domain.

FIX: just >touch NUM.h in ./num as first approximation

In the sources for 5.4.10, a very large file exists in dwtest/tone.wav. The file is about 170M in size!

This mirror does not have that file, because Github does not allow for the upload of files larger than 100M. You can download it yourself from the link above if you must.

Praat doesn't play the manipulation of the sound completely when the duration is manipulated to be above 3.

Steps to reproduce:

1. Get a Sound by whatever mean
2. Select the sound. Click "Manipulate - To manipulation"
3. Change the maximum value of range of duration to any value above 3 (e.g. 5)
4. Create a Duration point. Set the point to as high as possible so that the average duration becomes any length above 3.

Result:

• The audio isn't completely played in preview, and it doesn't completely exported after resynth.

I am trying to break audios into smaller parts on a linux server. But it requires opening an editor, which can't be done on batch. So is there a way to extract and save a specific portion of sound without opening the editor window, or a way to 'open' editor window while using praat from command line (batch)?

I've installed Praat on my ubuntu 16.04LTS using "sudo apt-get install praat".
After starting the programme two windows appear: "Pratt objects" and "Praat picture", but both of them don't have menu bars.

Since appending "-O2 -g -m64 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables" to CFLAGS in makefile.defs.linux.pulse, I'm seeing this:

I: Program causes undefined operation
   (likely same variable used twiceand post/pre incremented in the same expression).
   e.g. x = x++; Split it in two operations.
W: praat sequence-point Eigen.cpp:398

I: Program is likely to break with new gcc. Try -fno-strict-aliasing.
W: praat strict-aliasing-punning ../../sys/../num/../sys/melder.h:923
W: praat strict-aliasing-punning ../dwsys/../num/../sys/melder.h:923
W: praat strict-aliasing-punning ../num/../sys/melder.h:923
W: praat strict-aliasing-punning ../sys/../num/../sys/melder.h:923
W: praat strict-aliasing-punning Label.cpp:39

I: Statement is overflowing a buffer
E: praat bufferoverflow dictionary.cpp:3661:42

The last one is actually a problem with espeak, but since upstream hasn't fixed it yet (at least in stable), maybe you'll want to address it with this:

Index: praat-6.0.28/external/espeak/dictionary.cpp
===================================================================
--- praat-6.0.28.orig/external/espeak/dictionary.cpp
+++ praat-6.0.28/external/espeak/dictionary.cpp
@@ -3658,7 +3658,7 @@ int Lookup(Translator *tr, const char *w
                say_as = option_sayas;
                option_sayas = 0;   // don't speak replacement word as letter names
                text[0] = 0;
-               strncpy0(&text[1], word1, sizeof(text));
+               strncpy0(&text[1], word1, sizeof(text)-strlen(text)-1);
                found = TranslateWord(tr, &text[1], 0, NULL, NULL);
                strcpy(ph_out, word_phonemes);
                option_sayas = say_as;

Is there any reason to limit the number of accessible object attributes to the ones listed in the manual? It seems to me that in that list there is at least one omission, which is the time of the first sample (for sampled objects).

This was a (small) issue I bumped into while making a "To Table..." command for Pitch objects:

pitch = selected("Pitch")
# Is there any other way to obtain this value?
first = Get time from frame number: 1

matrix = To Matrix
transposed = Transpose
realtable = To TableOfReal

Insert column (index): 1
Set column label (index): 1, "time"
Set column label (index): 2, "f0"

# The time of the first sample is needed for this formula
Formula: "if col = 1 then row * Object_'pitch'.dx + 'first' else self fi"
table = To Table: "delete"
Remove column: "delete"
Rename: selected$("Table") - "_transposed"
removeObject: matrix, transposed, realtable
Formula (column range): "f0", "f0", "if !self then undefined else self fi"

I did the conversion almost entirely by using internal object casting commands, which I believe to be more efficient. But in order to turn the frame values of the Pitch object to seconds, the time of the first frame is needed.

In this case I could do this by getting that value with Get time from frame number... and then simply interpolating that value into the formula. But what I would have wanted to do was to make the formula

if col = 1 then row * Object_'pitch'.dx + Object_'pitch'.x1 else self fi

which is more transparent and easier to use in a formula that is not on a script. But this doesn't work, since x1 is not available as an attribute. In this case there was a workaround, but sice there is no command to obtain eg. the strength of each candidate in a Pitch object, the only way to read those is to write to text and parse the file externally, which seems silly.

In that case, what I would like to be able to write is

pitch = selected("Pitch")
frames = Get number of frames
for i to frames
  n = Object_'pitch'.frame[i].nCandidates
  for j to n
    strength[i][j] = Object_'pitch'.frame[i].candidate[j].strength
  endfor
endfor

What's the rationale for not allowing this?

It seems like this issue was fixed back in 2011, but as of the latest release (and it's been a while since I've installed praat, so I don't know how far back it goes), the tab shortcut is broken on linux. There is no way to play sound except with view>play...>(manually type in start and endpoint). Praat is unusable in this state, please rollback or otherwise fix as soon as possible.

With Praat 6.0.14 (64 bit, Xubuntu), I hit an issue this morning where the sound editor was not responding to mouse (click-drag to make selections) or keyboard (zooming commands like Ctrl+i, Ctrl+a, etc) commands. The commands were clearly registering, but the redraw to reflect the commands would not occur until I clicked outside the editor window (or pressed Alt+Tab) such that the editor window lost system focus (at which point if I had, say, click-dragged a selection and pressed Ctrl+n, the window would correctly show the waveform zoomed in to the selection I had made). The redraw would occur when the editor lost focus, not when focus returned to the editor. This happened only in one out of several sound editor windows this morning; closing that editor and re-opening from the same object in the list failed to reproduce. If it happens again, is there anything I should do / try to help narrow down the cause?

python:

import subprocess
subprocess.call(['D:\\Praat.exe', '--run', 'D:\\getInfo.praat', r'D:\a.wav'])

praat script:

form cmd_get_sound_info
    sentence file_path a.wav
endform
s = Read from file: file_path$
.....
.....
test1 View & Edit ----->not performed or completed
test2 Edit ----->not performed or completed
test3 editor:s ----->not performed or completed
.....
.....

Select... 0 2
startTime = Get start of selection
endTime = Get end of selection
name$ = selected$ ("Sound")

numberOfTimeSteps = (endTime - startTime) / 0.01
for step to numberOfTimeSteps
    time = startTime + (step - 1) * 0.01
    Move cursor to... time
    pitch = Get pitch
    intensity = Get intensity
    f1 = Get formant... 1
    f2 = Get formant... 2
    f3 = Get formant... 3
    f4 = Get formant... 4
    f5 = Get formant... 5

    appendInfoLine: fixed$ (time, 6), " ", fixed$ (pitch, 6)," ", fixed$ (intensity, 6)," ", fixed$ (f1, 6)," ", fixed$ (f2, 6)," ", fixed$ (f3, 6)," ", fixed$ (f4, 6)," ", fixed$ (f5, 6)
endfor

appendFile: name$+".txt", info$()

64-bit edition: praat6028_win64.zip
hope to get help,tks

Hi,

When adding a point to a PitchTier through the interface and then pasting the action history in the script window, the action is erroneously pasted as
Add point at: 0.0, 120

However, this will not run: the working command is
Add point: 0.0, 120

I'm used to confirming the quit dialog by just hitting the enter key.

This worked on OSX right up until v6.0.11.
However, from v6.0.12 onwards, hitting enter no longer works.

More critically however, any keyboard interaction, such as tabbing through elements in a form or dialog, is broken and causes Praat to crash (at least on 10.9.5).

According to the Manual (Scripting 6.9), running praat - should enter an interactive shell, or allow praat to accept commands from STDIN.
This worked up to v5.4.14:

$ /Volumes/Praat64_5414/Praat.app/Contents/MacOS/Praat -
Praat > echo Hello
Hello
Praat > Quit
$ 

However, from v5.4.15 onwards, this no longer works:

$ /Volumes/Praat64_5415/Praat.app/Contents/MacOS/Praat -
^C

It just hangs, and the Praat prompt never appears.

Hello! I hope that it is socially acceptable to post something like this here, I'm not yet a frequent user of GitHub.

I love using an external editor for scripting, and see no harm in adding a keyboard shortcut to the "Reopen from disk" command in the File menu of the Scripting editor. I cloned the repo, added this line, compiled, and it's working great for me. I hope you think it's a good idea to add this to the master branch? It's makes a keyboard shortcut of Shift + (command key/control key) + R.

sys/TextEditor.cpp line 606 is:

Editor_addCommand` (this, U"File", U"Reopen from disk", 0, menu_cb_reopen);

I suggest that it be:

Editor_addCommand (this, U"File", U"Reopen from disk", GuiMenu_SHIFT + 'R', menu_cb_reopen);

Again, I apologize if this isn't the place for this, but the change is so trivial I thought it wouldn't hurt to ask.

Dan

The command Line graph where... for Tables will not work if the column that is set to be the "vertical column" is the final one (=rightmost) in the Table object, and the Table object has fewer rows than columns.

This script exemplifies the problem:

# Generate sample Tables
columns$ = "a b c d e f g h i j k l m n o p q r s t u v w x y z"
Erase all
for rows from 3 to 5
  for cols from 3 to 5
    table = Create Table with column names: if rows < cols then "bad" else "good" fi,
      ... rows, left$(columns$, (2 * cols) - 1)

    # Populate Table
    for y to rows
      for x to cols
        Set numeric value: y, mid$(columns$, (2 * x) - 1), y
      endfor
    endfor

    # Problem only exists when drawing the final (=rightmost) column
    last_column$ = mid$(columns$, (2 * cols) - 1)

    # The problematic command
    Line graph where: last_column$, 0, 0, "", 0, 0, "+", 0, "yes", "1; (= everything)"

    # Problem only exists with Tables where nrow < ncol
    pauseScript: "Did ",
      ... if rows < cols then "not " else "" fi,
      ... "draw (", rows, ", ", cols, ")"
    removeObject: table
    Erase all
  endfor
endfor

The problem exists up to and including 5.4.19, in Linux and Mac at least.

Can you please write a rule for make install?

Praat is very strict about the data it handles, and will often refuse to open an object that is corrupt: if a string is expected and a number is found, for example.

However, this is not the case with all types of corruption. In particular, TextGrid files are very sensitive about their timestamps: the end of an interval must be the beginning of the interval that follows. And yet Praat will happily open a TextGrid in which this is not the case.

This leads to bugs that are difficult to track down. It would be better if Praat refused to open these files, maybe specifying where the error lies. Otherwise, opening any TextGrid leaves you with the possibility that Praat will behave unexpectedly under correct operation by the user, and there is no way to know this until it does.

Formula.h:22:10: fatal error: tensor.h: No such file or directory
 #include "tensor.h"

I see a sys/Tensor.h, is it the one?

With a sample script like:

form Test...
  real Number 1
endform

hitting "Run" will open a dialog box. Clicking "OK" will apply the changes and dismiss the box. Clicking "apply" will apply the changes and keep the dialog box open. If while a dialog box is open the script is run again, a new dialog box will appear and (at least in Linux) the previous dialog box will become greyed out.

Normal use can continue with the second dialog box. But closing the first one (which looks unresponsive) will crash Praat with a segmentation fault.

This is a particularly likely scenario when rapidly prototyping scripts.

I propose that, when the script is run, any existing forms are closed (as if by pressing "Cancel") before any new ones are open.

This happens systematically. I tested it with the Praat 6.0.15 binary from Github, running on a 32-bit machine under Kubuntu 15.10.

I notice that the PATH for system calls in praat scripts is this

/usr/bin:/bin:/usr/sbin:/sbin

on my system (OSX 10.9.5).

Is there a way to override/customize/extend the PATH for a system call? I would like to avoid having to specify the absolute path to the executable I want to run, to preserve cross-platform portability of the praat script.

More generally, can environment variables be modified (non-globally, one would hope) from within praat?

When extracting intervals (e.g. select audio file and corresponding TextGrid file -> Extract- -> Extract All Intervals...), labels such as '}', '@', and '@U' are changed to '', '', and '_U'.

Praat is absolutely my favourite speech analysis tool. I use it every day!

I've been trying to write a script to make lists of all files and directories under a given path. I've been doing that using recursive calls to script files (with runScript), and it's been working well (although maybe the problem is with the recursive script calls and not the Strings objects).

I came across two odd errors throwing segmentation faults or core dumps. And both seem to be triggered by lines in my script that would otherwise be pretty innocuous tasks:

1. Inserting a string with Insert string...
2. Removing a Strings object, both through the GUI or with commands like Remove or removeObject

The problems will come up for specific directory structures, but it's hard to predict which ones will be problematic. I don't think the problem is with the script, though, since the same script that will work in a directory with 2 subdirectories, crashes in a directory with 3 subdirectories. And the problem I find when removing the Strings object happens after my script has otherwise successfully finished.

Among the errors I get:

*** Error in `praat': free(): invalid next size (fast): 0x0e705d28 ***
Aborted (core dumped)

*** Error in `praat': corrupted double-linked list: 0x10148d20 ***
Aborted (core dumped)

*** Error in `praat': double free or corruption (out): 0x0edcd830 ***
Aborted (core dumped)

Other times, I just get a segmentation fault.

The problems are erratic, in that they appear only when querying some directories (with specific sub-directory structures), and sometimes only when using specific values for the max_depth argument, which modifies how deep the recursion should be. In these cases, running the script with a max_depth of 0 throws no errors, but specifying a value of eg. 2 might give me the double free or corruption error above.

This I tested on a machine running Kubuntu 14.04 (32-bit) using Praat 5.4.22. The problems exist both from the command line and from the GUI.

On OSX 10.9.5, I've noticed that the way the labels are displayed in intervals of a TextGrid Editor has weird overprinting and line breaks (depending on window size, zoom level, interval length, and label).

In v5.4.11 through v5.4.15, it looks like this:

But in v5.4.10 and before it was fine:

The help page is missing in 6.0.15, and has been missing since at least 5.4.22, at least for the Get shimmer (local)... command.

The help page is missing in at least 6.0.15 for all other versions of Get shimmer.

Selecting "Help" in the Scale intensity... command results in an error.

Running Praat 6.0.04 on Kubuntu 14.04.

Console Praat (praatcon.exe) 5.4.21 (September 29 Release) is crashing on Windows 8, and Windows 7 (those are the only ones I have checked with)

A sample script that generates the waveform and saves that as a PNG file is attached.

Unable to attach script due to some issue with Github interface, I am copy-pasting it below

#praat test file , Please change the following two paths
outputFolder$ = "C:\Temp\del\praat"
audioFile$ = "C:\Temp\del\praat\test.mp3"

if audioFile$ <> ""
    Read from file... 'audioFile$'
    audioName$ = selected$ ("Sound")
endif

Line width... 1
Viewport... 0 4 0 1
Colour... magenta
Draw... 0 0 0 0 no curve
Draw inner box

select Sound 'audioName$'
To Intensity... 80 0
Down to IntensityTier

select Sound 'audioName$'
To Pitch (ac)... 0.0 80 15 no 0.03 0.3 0.2 0.9 0.14 300
Down to PitchTier

select Intensity 'audioName$'
Viewport... 0 4 0.6 1.6

Colour... lime
Line width... 2
Draw... 0 0 0 0 no
#Text right... yes Intensity (dBs)

# Draw pitch track
select Pitch 'audioName$'
Viewport... 0 4 0.6 1.6

# Draw in either greyscale or color

Colour... blue

Line width... 1
Draw... 0 0 80 300 yes
Marks bottom every... 1 5 yes yes no
Line width... 2
Draw... 0 0 80 300 no

Viewport... 0 4 0 1.6
Save as 300-dpi PNG file... 'outputFolder$'\waveform.png

as reported by rpmlint:

praat.x86_64: W: incorrect-fsf-address /usr/share/doc/packages/praat/GNU_General_Public_License.txt
The Free Software Foundation address in this file seems to be outdated or
misspelled.  Ask upstream to update the address, or if this is a license file,
possibly the entire file with a new copy available from the FSF.

praat.x86_64: I: binary-or-shlib-calls-gethostbyname /usr/bin/praat
The binary calls gethostbyname(). Please port the code to use getaddrinfo().

When writing procedures, I often find it useful to differentiate between procedures for internal use, and procedures for general use. Taking a hint from Perl and others, I like to preppend internal procedures with an initial underscore (_).

Variables in Praat are not allowed to begin with an underscore. But since "local" variables in procedures take their name from the name of the procedure they belong to, and procedures can be named with an initial underscore, this results in all sorts of fun with Praat variables:

procedure _test ()
  .local = 1
endproc
@_test()

# Works
a = '_test.local'
appendInfoLine: a

# Doesn't
asserterror Unknown symbol:'newline$'« _
a = _test.local

Is this desired behaviour? I've learned now to navigate around the problem as illustrated above, but it feels a bit inconsistent. This is one of the reasons I sometimes still use variable substitution.

Script is attached with this issue. We can share the audio that is causing the problem if need be. We ran the script through command line as well as the UI, we did see the Praat Picture fine, but the Save as PNG File... instruction is creating 0 bytes file. Any help on this will be appreciated.

You may have to correct the paths in the attached script
waveform.praat.txt

I feel this has been a long lasting bug, but seeing that the display of man pages seems to have been improved recently, I figured it made sense to add this in.

At least in Linux (tested on other systems, but confirmed in Kubuntu 14.04 with Praat 5.4.15), switching in the manual from a standard font to an italic font causes the italic to be printed on top of the end of the string in a normal font. Sample image below:

At its most serious, this can make some change-heavy passages impossible to understand, meaning users need to refer to the HTML documentation.

I am generating the waveform from praat script. It's working great but only for attached audio file https://drive.google.com/file/d/0B820TayaGz1PUjhxUTlIVzZRQmM/view?usp=sharing for which I have a problem.

Below script generates the multiple waveform images based on its duration since there is a problem creating a single waveform for the longer audio so we have chosen this option to create waveforms for longer audio. This script generally working great for many audio files but we have got the problem for the audio https://drive.google.com/file/d/0B820TayaGz1PUjhxUTlIVzZRQmM/view?usp=sharing. It's generating 0 byte waveform that means no contents on waveform image while we are expecting the file size > 1 KB.

To test this audio please download the audio file from the above link and change the values for variables outputFolder$ (output location for waveform) and audioFile$ (source path of the audio file). After changing the values then please run the script from the praat application.

We have confirmed that there is a problem while saving waveform images whose responsible code is in bold text in the below script.

Below is the script which I am using.

outputFolder$ = "C:\Temp\ffmpeg\experiments"
audioFile$ = "D:\audio.mp3"

window_start = 0
full_window=3600;
divide=15;

if audioFile$ <> ""
Read from file... 'audioFile$'
audioName$ = selected$ ("Sound")
endif
To TextGrid... phrases
select TextGrid 'audioName$'
file_length = Get total duration

if file_length<=41*60
full_window=2460
divide=30;
endif

if file_length<=21*60
full_window=1260
divide=20;
endif

if file_length<=11*60
full_window=660
divide=15;
endif

if file_length>41*60
full_window=3600
divide=60;
endif

draw_num = floor(file_length/full_window)

last_draw = file_length - draw_num*full_window

if file_length<full_window
view_width = 4 * file_length/divide
else
view_width = 4 * full_window/divide
endif
view_width=view_width

for b from 1 to draw_num
select TextGrid 'audioName$'
window_end = window_start + full_window
start_int = Get interval at time... 1 window_start
end_int = Get interval at time... 1 window_end
select Sound 'audioName$'
Line width... 1
Viewport... 0 view_width 0 2
Colour... {110, 110, 10}
Draw... window_start window_end 0 0 no curve

select Sound 'audioName$'
To Intensity... 80 0
Down to IntensityTier

select Sound 'audioName$'
To Pitch (ac)... 0.0 80 15 no 0.03 0.3 0.2 0.9 0.14 300
Down to PitchTier

select Intensity 'audioName$'
Viewport... 0 view_width 1.25 2.25

Colour... lime
Line width... 2

select Pitch 'audioName$'
Viewport... 0 view_width 1.25 2.25

Colour... green

Line width... 1
Draw... window_start window_end 80 300 no
Line width... 2
Draw... window_start window_end 80 300 no

save_width = view_width - 0.60
Viewport... 0.60 save_width 0.40 2
Save as 300-dpi PNG file... 'outputFolder$'\waveform_'b'.png
Erase all
window_start = window_start + full_window
endfor

select TextGrid 'audioName$'
window_start = file_length - last_draw
window_end = file_length
start_int = Get interval at time... 1 window_start
end_int = Get interval at time... 1 window_end
view_width = 4 * (window_end-window_start)/divide
select Sound 'audioName$'
Line width... 1
Viewport... 0 view_width 0 2
Colour... {110, 110, 10}
Draw... window_start window_end 0 0 no curve

select Sound 'audioName$'
To Intensity... 80 0
Down to IntensityTier

select Sound 'audioName$'
To Pitch (ac)... 0.0 80 15 no 0.03 0.3 0.2 0.9 0.14 300
Down to PitchTier

select Intensity 'audioName$'
#Viewport... 0 view_width 1.25 2.75
Viewport... 0 view_width 1 2.75

Colour... lime
Line width... 2

select Pitch 'audioName$'
Viewport... 0 view_width 1.25 2.75
Colour... green

Line width... 1
Draw... window_start window_end 80 300 no
Line width... 2
Draw... window_start window_end 80 300 no

final_num = draw_num + 1
save_width = view_width - 0.60
Viewport... 0.60 save_width 0.40 2.75
Save as 300-dpi PNG file... 'outputFolder$'\waveform.png
Erase all
window_start = window_start + full_window

Instead of using --run, it would be nice to have a way to directly feed scripts to a running console-mode version of the executable. Using such an approach, a script could open the executable once, pipe commands to it, and send a null character to flush the buffer. This would make using Praat from an external script (such as with PraatR) much faster.

It would be nice to have access to the functionality of Praat not just from the command line, but from a C library that could be used directly, or wrapped into other libraries, such as PraatR.

In the following code snippet, no text will be printed in the Info window:

clearinfo
file$ = temporaryDirectory$ + "/temp.praat"
writeFileLine: file$,
  ... "writeInfoLine: ""Hello"""
a = nocheck runScript: file$
deleteFile(file$)

However, if the nocheck directive is removed,

a = runScript: file$

or if the line with the call to runScript is no longer an assignment

nocheck runScript: file$

the Info window correctly displays the string "Hello". Is this intentional? If so, what's the rationale?

Hello,

Is it possible to cross compile Praat for Android ? I only need the command line version without any graphics and sound packages. I only want to execute basic praat script without user interaction.

Best regards,

Jimmy

Normally, it's not a problem that variables and functions have the same names (which is definitely a good thing):

string$ = "hello"
appendInfoLine: string$

index = 1
appendInfoLine: index

But indexed variables seem to be an exception:

string$[1] = "hello"
asserterror Expected the end of the formula, but found "["
appendInfoLine: string$[1]

right$[1] = "hello"
asserterror Unknown variable:
appendInfoLine: right$[1]

# Same with numeric functions
index[1] = 1
asserterror Unknown variable:
appendInfoLine: index[1]

Is this a desired behaviour? I would understand if variables could never be named like functions. But the fact that some of them can, and some of them cannot, is confusing.

It's also confusing that the error is raised when the variable is used, and not when it is assigned. If those variable names are invalid, then surely the error should be caught sooner?

With recent versions of Praat, I cannot play audio on Kubuntu 14.04 (32 bit). Reverting back to 5.4.22 I encounter no such problem. The problem exists in Praat 6.0 and above (6.0.04 included).

The error that pops up tells me to turn on PulseAudio in the Sound Playing Preferences window, but PulseAudio is already selected. Selecting "Alsa via PulseAudio" fixes the problem during the session, but that preference is reverted when restarting Praat, so it needs to be set again each time.

rpmlint complains about this:

I: Program returns random data in a function
E: praat no-return-in-nonvoid-function TextGrid_extensions.cpp:817

This is my build command (on openSUSE linux):

cp makefiles/makefile.defs.linux.pulse ./makefile.defs
+ sed -e '/^CFLAGS/s/$/\ -fmessage-length=0 -grecord-gcc-switches -O2 -Wall -D_FORTIFY_SOURCE=2 -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables/' -i makefile.defs
+ make -j4