A customer wanted me to use SAPI, but also wanted events from SAPI to fire. Google says nobody has been successful at getting SAPI's events to work in Perl. That is because of a 3 years and counting, patch provided, bug rt://43574 in Win32::OLE. Once fixed, the example below works. 
use strict;
use Win32::OLE qw(EVENTS);
use Win32::OLE::Const;
use Data::Dumper;

#Win32::OLE::Const is terrible, the tree like interface ( Win32::OLE::
+TypeInfo class )
#is explictly undocumented
my $constHash = Win32::OLE::Const->Load('Microsoft Speech Object Libra
+ry');

#print Dumper($constHash);

my $vox = Win32::OLE->new ('SAPI.SpVoice')
 || die "Unable to create SAPI object\n";

sub Event {
    my %hash;
    #foreach (@_) {
    #    print " ".$_." ";
    #}
    #print "\n";
    #system("pause");
#stack trace from the system pause above
#     ntdll.dll!_KiFastSystemCallRet@0()     
#     ntdll.dll!_ZwWaitForMultipleObjects@20()  + 0xc    
#     kernel32.dll!_WaitForMultipleObjectsEx@20()  - 0x48    
#     user32.dll!_RealMsgWaitForMultipleObjectsEx@20()  + 0xd9    
#     user32.dll!_MsgWaitForMultipleObjects@20()  + 0x1f    
#>    perl512.dll!win32_msgwait(interpreter * my_perl=0x003940dc, unsi
+gned long count=0x00000001, void * * handles=0x0006f500, unsigned lon
+g timeout=0xffffffff, unsigned long * resultp=0x00000000)  Line 2181 
++ 0x19    C
#     perl512.dll!win32_spawnvp(int mode=0x00000000, const char * cmdn
+ame=0x00a11604, const char * const * argv=0x00a1abfc)  Line 4249 + 0x
+19    C
#     perl512.dll!do_spawn2(interpreter * my_perl=0x003940dc, const ch
+ar * cmd=0x0083a3f4, int exectype=0x00000002)  Line 784 + 0x11    C
#     perl512.dll!Perl_do_spawn(interpreter * my_perl=0x003940dc, char
+ * cmd=0x0083a3f4)  Line 820 + 0xf    C
#     perl512.dll!Perl_pp_system(interpreter * my_perl=0x003940dc)  Li
+ne 4277 + 0x5e    C
#     perl512.dll!Perl_runops_debug(interpreter * my_perl=0x003940dc) 
+ Line 2049 + 0xd    C
#     perl512.dll!Perl_call_sv(interpreter * my_perl=0x003940dc, sv * 
+sv=0x00a042ec, volatile long flags=0x00000006)  Line 2590 + 0x36    C
#     OLE.dll!EventSink::Invoke(long dispidMember=0x00000003, const _G
+UID & riid={...}, unsigned long lcid=0x00000400, unsigned short wFlag
+s=0x0001, tagDISPPARAMS * pdispparams=0x0006f7d8, tagVARIANT * pvarRe
+sult=0x00000000, tagEXCEPINFO * pexcepinfo=0x00000000, unsigned int *
+ puArgErr=0x00000000)  Line 2161 + 0xf    C++
#     sapi.dll!CProxy_ISpeechVoiceEvents<CSpVoice>::Fire_VoiceChange()
+  + 0xb8    
#     sapi.dll!CSpVoice::NotifyCallback()  + 0x11d    
#     sapi.dll!CSpNotify::WndProc()  + 0x60    
#     user32.dll!_InternalCallWinProc@20()  + 0x28    
#     user32.dll!_UserCallWinProcCheckWow@32()  + 0xb7    
#     user32.dll!_DispatchMessageWorker@8()  + 0xdc    
#     user32.dll!_DispatchMessageA@4()  + 0xf    
#     OLE.dll!XS_Win32__OLE_Initialize(interpreter * my_perl=0x003940d
+c, cv * cv=0x008d0634)  Line 3549    C++
#     perl512.dll!Perl_pp_entersub(interpreter * my_perl=0x003940dc)  
+Line 2882 + 0x10    C
#     perl512.dll!Perl_runops_debug(interpreter * my_perl=0x003940dc) 
+ Line 2049 + 0xd    C
#     perl512.dll!S_run_body(interpreter * my_perl=0x003940dc, long ol
+dscope=0x00000001)  Line 2308 + 0xd    C
#     perl512.dll!perl_run(interpreter * my_perl=0x003940dc)  Line 223
+3 + 0xd    C
#     perl512.dll!RunPerl(int argc=0x00000002, char * * argv=0x0028251
+0, char * * env=0x002828f0)  Line 270 + 0x9    C++
#     perl.exe!main(int argc=0x00000002, char * * argv=0x00282510, cha
+r * * env=0x00282d40)  Line 23 + 0x12    C
#     perl.exe!mainCRTStartup()  Line 398 + 0xe    C
#     kernel32.dll!_BaseProcessStart@4()  + 0x23    

    $hash{'eventName'} = $_[1];
    if($hash{'eventName'} eq 'AudioLevel'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
        $hash{'AudioLevel'} = $_[4];
    }
    elsif($hash{'eventName'} eq 'Phoneme'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
        $hash{'Duration'} = $_[4];
        $hash{'NextPhoneId'} = $_[5];
        $hash{'Feature'} = DecodeSpeechVisemeFeature($_[6]);
        $hash{'CurrentPhoneId'} = $_[7];
    }
    elsif($hash{'eventName'} eq 'Word'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
        $hash{'CharacterPosition'} = $_[4];
        $hash{'Length'} = $_[5];
    }
    elsif($hash{'eventName'} eq 'StartStream'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
    }
    elsif($hash{'eventName'} eq 'VoiceChange'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
        #Dumping the obj is pagefuls of info
        $hash{'VoiceObjectTokenDesc'} = $_[4]->GetDescription();
    }
    elsif($hash{'eventName'} eq 'Sentence'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
        $hash{'CharacterPosition'} = $_[4];
        $hash{'Length'} = $_[5];
    }
    elsif($hash{'eventName'} eq 'Viseme'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
        $hash{'Duration'} = $_[4];
        $hash{'NextVisemeId'} = DecodeSpeechVisemeType($_[5]);
        $hash{'Feature'} = DecodeSpeechVisemeFeature($_[6]);
        $hash{'CurrentVisemeId'} = DecodeSpeechVisemeType($_[7]);
    }
    elsif($hash{'eventName'} eq 'EndStream'){
        $hash{'StreamNumber'} = $_[2];
        $hash{'StreamPosition'} = $_[3];
    }
    else {die "unknown event $hash{'eventName'}";}

    print Dumper(\%hash);
    if($hash{'eventName'} eq 'EndStream'){
        Win32::OLE->QuitMessageLoop();
    }
}

my $res = Win32::OLE->WithEvents ($vox, \&Event
                                  #'_' bug in Win32 OLE, must use UUID
+ if not fixed
                                  #https://rt.cpan.org/Public/Bug/Disp
+lay.html?id=43574
                                  #, '{A372ACD1-3BEF-4BBD-8FFB-CB3E2B4
+16AF8}'
                                  ,'_ISpeechVoiceEvents'
                                  );
print " OLE last error ".Win32::OLE->LastError()."\n";
#print Dumper($res);

die "bad const" if $constHash->{'SVEAllEvents'}  !=  33790;
$vox->{'EventInterests'} = $constHash->{'SVEAllEvents'}; 
$res = $vox->{'EventInterests'};
print "EventInterests res ".Dumper($res);

my $text = "I Love Perl";

die "bad const" if $constHash->{'SVSFlagsAsync'}  != 1;
my $res = $vox->Speak ($text, $constHash->{'SVSFlagsAsync'});
print Dumper($res);


print " ".Win32::OLE->LastError()."\n";
print "WFNE ".Dumper($res);
#potential ideas, investigate using Perl's alarm win msg loop
Win32::OLE->MessageLoop();

{
my @table;
$table[0] = 'silence';
$table[1] = 'ae ax ah';
$table[2] = 'aa';
$table[3] = 'ao';
$table[4] = 'ey eh uh';
$table[5] = 'er';
$table[6] = 'y iy ih ix';
$table[7] = 'w uw';
$table[8] = 'ow';
$table[9] = 'aw';
$table[10] = 'oy';
$table[11] = 'ay';
$table[12] = 'h';
$table[13] = 'r';
$table[14] = 'l';
$table[15] = 's z';
$table[16] = 'sh ch jh zh';
$table[17] = 'th dh';
$table[18] = 'f v';
$table[19] = 'd t n';
$table[20] = 'k g ng';
$table[21] = 'p b m';
sub DecodeSpeechVisemeType {
    return $table[$_[0]];
}
}
sub DecodeSpeechVisemeFeature {
        if($_[0] == 0){return 'None';}
        elsif($_[0] == 1){return 'Stressed';}
        elsif($_[0] == 2){return 'Emphasis';}
        else{die "unknown SpeechVisemeFeature";}
}

[download]
The output of the script. 
 OLE last error 0
EventInterests res $VAR1 = 33790;
$VAR1 = 1;
 0
WFNE $VAR1 = 1;
$VAR1 = {
          "StreamNumber" => 1,
          "StreamPosition" => 0,
          "eventName" => "StartStream"
        };
$VAR1 = {
          "StreamNumber" => 1,
          "StreamPosition" => 0,
          "VoiceObjectTokenDesc" => "LH Michael",
          "eventName" => "VoiceChange"
        };
$VAR1 = {
          "CharacterPosition" => 0,
          "Length" => 22,
          "StreamNumber" => 1,
          "StreamPosition" => 0,
          "eventName" => "Sentence"
        };
$VAR1 = {
          "CharacterPosition" => 0,
          "Length" => 1,
          "StreamNumber" => 1,
          "StreamPosition" => 0,
          "eventName" => "Word"
        };
$VAR1 = {
          "CurrentPhoneId" => 0,
          "Duration" => 15,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 0,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 15,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 0,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "CurrentPhoneId" => 1,
          "Duration" => 15,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 330,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 15,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 330,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "CurrentPhoneId" => 11,
          "Duration" => 145,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 660,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 145,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 660,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 31,
          "StreamNumber" => 1,
          "StreamPosition" => 2261,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "CharacterPosition" => 2,
          "Length" => 4,
          "StreamNumber" => 1,
          "StreamPosition" => 3850,
          "eventName" => "Word"
        };
$VAR1 = {
          "CurrentPhoneId" => 14,
          "Duration" => 90,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 3850,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 90,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 3850,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 28,
          "StreamNumber" => 1,
          "StreamPosition" => 4649,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "CurrentPhoneId" => 1,
          "Duration" => 94,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 5830,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 94,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 5830,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "CurrentPhoneId" => 18,
          "Duration" => 63,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 7898,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 63,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 7898,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 17,
          "StreamNumber" => 1,
          "StreamPosition" => 8007,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "CharacterPosition" => 7,
          "Length" => 4,
          "StreamNumber" => 1,
          "StreamPosition" => 9284,
          "eventName" => "Word"
        };
$VAR1 = {
          "CurrentPhoneId" => 21,
          "Duration" => 121,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 9284,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 121,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 9284,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 5,
          "StreamNumber" => 1,
          "StreamPosition" => 11111,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "CurrentPhoneId" => 4,
          "Duration" => 155,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 11946,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 155,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 11946,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 28,
          "StreamNumber" => 1,
          "StreamPosition" => 13493,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "CurrentPhoneId" => 13,
          "Duration" => 90,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 15356,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 90,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 15356,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 27,
          "StreamNumber" => 1,
          "StreamPosition" => 16835,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "CurrentPhoneId" => 14,
          "Duration" => 101,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 17336,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 101,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 17336,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "CurrentPhoneId" => 0,
          "Duration" => 421,
          "Feature" => "Emphasis",
          "NextPhoneId" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 19558,
          "eventName" => "Phoneme"
        };
$VAR1 = {
          "CurrentVisemeId" => undef,
          "Duration" => 421,
          "Feature" => "Emphasis",
          "NextVisemeId" => undef,
          "StreamNumber" => 1,
          "StreamPosition" => 19558,
          "eventName" => "Viseme"
        };
$VAR1 = {
          "AudioLevel" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 20169,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "AudioLevel" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 23507,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "AudioLevel" => 0,
          "StreamNumber" => 1,
          "StreamPosition" => 26847,
          "eventName" => "AudioLevel"
        };
$VAR1 = {
          "StreamNumber" => 1,
          "StreamPosition" => 28820,
          "eventName" => "EndStream"
        };

[download]

Replies are listed 'Best First'.
Re: SAPI and Win32::OLE's broken WithEvents
by Tanktalus (Canon) on May 07, 2012 at 18:13 UTC

    A couple of comments on the code.

    First, in Events, you have a massive if/elsif/.../else. And every item is checking the same thing - the event name. This looks like a perfect use for given/when:

    given ($hash{'eventName'})
{
  when ('AudioLevel') {...}
  ...
}

    [download]
    Except that you're doing the same thing in each clause (except for the default/unknown case). So a hash of arrays would likely work simpler: 
    my %keys = (
  AudioLevel => [qw(StreamNumber StreamPosition AudioLevel)],
  Phoneme    => [qw(StreamNumber StreamPosition Duration NextPhoneId F
+eature CurrentPhoneId)],
  # ...
);
if ($keys{$hash{'eventName'}})
{
  @hash{@{$keys{$hash{'eventName'}}}} = @_[2..$#_];
}
else
{
  die "unknown event $hash{'eventName'}";
}

    [download]
    I think that'll produce more concise and, more importantly, more readable code. Of course, it's missing important bits where you call other functions, and this is definitely solvable (especially with anonymous code refs in your %keys hash, assuming that the assignment code here looks for them and calls them). You may prefer given/when as it's less of a change from how you're currently thinking, but I generally prefer code refs. 
    my %keys = (
  ...
  Viseme => [ qw(StreamNumber StreamPosition Duration),
              {
                NextVisemeId => sub { DecodeSpeechVisemeType($_[5]) },
                Feature => sub { DecodeSpeechVisemeFeature($_[6]) },
                CurrentVisemeId => sub { DecodeSpeechVisemeType($_[7])
              }
            ],
  ...
);

if ($keys{$hash{'eventName'}})
{
  my @keys = @{$keys{$hash{'eventName'}}};
  my $dynamic_keys = ref $keys[-1] ? pop @keys : {};

  @hash{@keys} = @_[2..$#keys+2];

  for my $dyn (keys %$dynamic_keys)
  {
    $hash{$dyn} = $dyanmic_keys->{$dyn}->(@_);
  }
}
else
{
  die "unknown event $hash{'eventName'}";
}

    [download]
    I don't think that's too bad. A bit more difficult than if nothing was dynamic, but that's kind of the price you pay for that dynamicism.

    Second is your @table for the DecodeSpeechVisemeType function. You may be better off putting the @table inside your function and calling it a state variable instead of a my. You can still initialise it once this way, but gets rid of some extra braces. But it also means that you need to learn to initialise arrays as part of their declaration. Instead of my @table; $table[0] =..., say my @table = ( 'silence', 'ae ax ah', ... );. I'm actually surprised what you have is working, because my understanding is that your initialisation for @table wouldn't be called until after the Win32::OLE->MessageLoop() returns.

    Finally, DecodeSpeechVisemeFeature - again with the if/elsif/else. Use a lookup hash. Or, in this case, a lookup array.

    Warning: all code above untested. :-)

[reply]
[d/l]
[select]
      I'm actually surprised what you have is working, because my understanding is that your initialisation for @table wouldn't be called until after the Win32::OLE->MessageLoop() returns.

      Your right about @table. Its broken, in the output its all undef. I failed at the closure. Should have put a BEGIN block around it.

      @hash{@{$keys{$hash{'eventName'}}}} = @_2..$#_;

      "@hash{" what is that? hash is a % or a $ for slices.
[reply]
[reply]
Re: SAPI and Win32::OLE's broken WithEvents
by Anonymous Monk on May 05, 2012 at 23:54 UTC

    And the meditation part is?

[reply]

Back to Meditations