test_festival

// preprocessor directives
#include <festival.h>
#include <sys/time.h>
using namespace std;



//
EST_Val default_val_float(0.0f);



// functions for accessing word start & end times in an utterance tree

// returns segment start time
EST_Val ff_seg_start(EST_Item* s)
{
  EST_Item* n = as(s, "Segment");
  if (prev(n) == 0) return default_val_float;
  return prev(n)->F("end", 0);
} // ff_seg_start(EST_Item*)



// returns segment end time
EST_Val ff_seg_end(EST_Item* s)
{
  return s->F("end", 0);
} // ff_seg_end(EST_Item*)



// returns syllable end time
EST_Val ff_syl_end(EST_Item* s)
{
  EST_Item* n = as(s, "SylStructure");
  if (daughtern(n) == 0) return default_val_float;
  return ff_seg_end(daughtern(n));
} // ff_syl_end(EST_Item*)



// returns word start time
EST_Val ff_word_start(EST_Item* s)
{
  EST_Item* n = as(s, "SylStructure");
  if (daughter1(daughter1(n)) == 0) return default_val_float;
  return ff_seg_start(daughter1(daughter1(n)));
} // ff_word_start(EST_Item*)



// returns word end time
EST_Val ff_word_end(EST_Item* s)
{
  EST_Item* n = as(s, "SylStructure");
  if (daughtern(n) == 0) return default_val_float;
  return ff_syl_end(daughtern(n));
} // ff_word_end(EST_Item*)



// executes main program code
int main(int argc, char** argv)
{
  if (argc < 2)
  {
    printf(">> ERROR: Invalid arguments...\n\n");
    return 1;
  }

  // Festival local variables
  int heap_size       = 210000;  // default scheme heap size
  int load_init_files = 1;       // we want the festival init files loaded

  // initialize the Festival text-to-speech (TTS) engine
  festival_initialize(load_init_files, heap_size);


  // iterate through text input; append words to string str
  EST_String str;
  for(int i = 1; i < argc; ++i)
  {
    //printf("arg %d: %s\n", i, argv[i]);
    str += argv[i];
    str += " ";
  }

  // Lisp command to create and synthesize a text utterance with the user input
  EST_String command1 = "(set! utt1 (utt.synth(Utterance Text \"" +
                        str + "\")))";

  // Lisp command to save the utterance to a file
  EST_String command2 = "(utt.save utt1 \"inputText.utt\")";

  // evaluate the Lisp command
  bool success1 = festival_eval_command(command1);
  bool success2 = festival_eval_command(command2);
  //cout << "Command[1] success: " << success1 << endl;
  //cout << "Command[2] success: " << success2 << endl;

  // load the utterance previously made
  EST_Utterance myUtt;
  myUtt.load("inputText.utt");

  // print out how many words are in this utterance
  EST_Relation myRel = *myUtt.relation("Word");
  cout << "# words: " << myRel.length() << endl;

  // iterate through the utterance tree to get each word
  EST_Item* s = NULL;
  for (s = myUtt.relation("Word")->head(); s != 0; s = next(s))
  {
    //print out features of the word: name, start time, end time
    cout << "Word: "   << s->S("name");
    cout << "\tstart: "<< ff_word_start(s);
    cout << "\tend: "  << ff_word_end(s) << endl;

    // print out the "word_end" feature of the word  <-- does not work
    //cout << s->A("word_end") << endl;
  }

  // synthesize the parameterized text and time its duration
  timeval t_start, t_stop;
  gettimeofday(&t_start, NULL);
  festival_say_text(str);
  gettimeofday(&t_stop, NULL);
  double t_elapsed = (t_stop.tv_sec  + (double)t_stop.tv_usec  * 1e-6) -
                     (t_start.tv_sec + (double)t_start.tv_usec * 1e-6);
  printf("Time elapsed: %.6f\n", t_elapsed);

  return 0;
} // main(int, char**)