CPAN::Groonga

package Speech::Recognition::Vosk;
use strict;

our $VERSION = '0.01';

=head1 NAME

Speech::Recognition::Vosk - offline voice recognition using the Vosk toolkit

=head1 SYNOPSIS

Most likely, you want to use the more convenient OO wrapper in
L<Speech::Recognition::Vosk::Recognizer>.

  use Speech::Recognition::Vosk;
  use JSON 'decode_json';

  my $model = Speech::Recognition::Vosk::model_new("model-en");
  my $recognizer = Speech::Recognition::Vosk::recognizer_new($model, 44100);

  binmode STDIN, ':raw';

  while( ! eof(*STDIN)) {
      read(STDIN, my $buf, 3200);
      my $complete = Speech::Recognition::Vosk::recognizer_accept_waveform($recognizer, $buf);
      my $spoken;
      if( $complete ) {
          $spoken = Speech::Recognition::Vosk::recognizer_result($recognizer);
      } else {
          $spoken = Speech::Recognition::Vosk::recognizer_partial_result($recognizer);
      }

      my $info = decode_json($spoken);
      if( $info->{text}) {
          print $info->{text},"\n";
      } else {
          local $| = 1;
          print $info->{partial}, "\r";
      };
  }

  # Flush the buffers
  my $spoken = Speech::Recognition::Vosk::recognizer_final_result($recognizer);
  my $info = decode_json($spoken);
  print $info->{text},"\n";

=head1 FUNCTIONS

=cut

require XSLoader;
XSLoader::load(__PACKAGE__, $VERSION);

1;