Source Code for Module synthDrivers._espeak

  1  #synthDrivers/_espeak.py 
  2  #A part of NonVisual Desktop Access (NVDA) 
  3  #Copyright (C) 2006-2007 NVDA Contributors <http://www.nvda-project.org/> 
  4  #This file is covered by the GNU General Public License. 
  5  #See the file COPYING for more details. 
  6   
  7  import time 
  8  import nvwave 
  9  import threading 
 10  import Queue 
 11  from ctypes import * 
 12  import config 
 13  import globalVars 
 14  from logHandler import log 
 15  import os 
 16  import codecs 
 17   
 18  isSpeaking = False 
 19  lastIndex = None 
 20  bgThread=None 
 21  bgQueue = None 
 22  player = None 
 23  espeakDLL=None 
 24   
 25  #Parameter bounds 
 26  minRate=80 
 27  maxRate=450 
 28  minPitch=0 
 29  maxPitch=99 
 30   
 31  #event types 
 32  espeakEVENT_LIST_TERMINATED=0 
 33  espeakEVENT_WORD=1 
 34  espeakEVENT_SENTENCE=2 
 35  espeakEVENT_MARK=3 
 36  espeakEVENT_PLAY=4 
 37  espeakEVENT_END=5 
 38  espeakEVENT_MSG_TERMINATED=6 
 39  espeakEVENT_PHONEME=7 
 40   
 41  #position types 
 42  POS_CHARACTER=1 
 43  POS_WORD=2 
 44  POS_SENTENCE=3 
 45   
 46  #output types 
 47  AUDIO_OUTPUT_PLAYBACK=0 
 48  AUDIO_OUTPUT_RETRIEVAL=1 
 49  AUDIO_OUTPUT_SYNCHRONOUS=2 
 50  AUDIO_OUTPUT_SYNCH_PLAYBACK=3 
 51   
 52  #synth flags 
 53  espeakCHARS_AUTO=0 
 54  espeakCHARS_UTF8=1 
 55  espeakCHARS_8BIT=2 
 56  espeakCHARS_WCHAR=3 
 57  espeakSSML=0x10 
 58  espeakPHONEMES=0x100 
 59  espeakENDPAUSE=0x1000 
 60  espeakKEEP_NAMEDATA=0x2000 
 61   
 62  #speech parameters 
 63  espeakSILENCE=0 
 64  espeakRATE=1 
 65  espeakVOLUME=2 
 66  espeakPITCH=3 
 67  espeakRANGE=4 
 68  espeakPUNCTUATION=5 
 69  espeakCAPITALS=6 
 70  espeakWORDGAP=7 
 71  espeakOPTIONS=8   # reserved for misc. options.  not yet used 
 72  espeakINTONATION=9 
 73  espeakRESERVED1=10 
 74  espeakRESERVED2=11 
 75   
 76  #error codes 
 77  EE_OK=0 
 78  #EE_INTERNAL_ERROR=-1 
 79  #EE_BUFFER_FULL=1 
 80  #EE_NOT_FOUND=2 
 81   
 82 -class espeak_EVENT_id(Union): 
 83          _fields_=[ 
 84                  ('number',c_int), 
 85                  ('name',c_char_p), 
 86          ] 
 87   
 88 -class espeak_EVENT(Structure): 
 89          _fields_=[ 
 90                  ('type',c_int), 
 91                  ('unique_identifier',c_uint), 
 92                  ('text_position',c_int), 
 93                  ('length',c_int), 
 94                  ('audio_position',c_int), 
 95                  ('sample',c_int), 
 96                  ('user_data',c_void_p), 
 97                  ('id',espeak_EVENT_id), 
 98          ] 
 99   
100 -class espeak_VOICE(Structure): 
101          _fields_=[ 
102                  ('name',c_char_p), 
103                  ('languages',c_char_p), 
104                  ('identifier',c_char_p), 
105                  ('gender',c_byte), 
106                  ('age',c_byte), 
107                  ('variant',c_byte), 
108                  ('xx1',c_byte), 
109                  ('score',c_int), 
110                  ('spare',c_void_p), 
111          ] 
112   
113 -        def __eq__(self, other): 
114                  return isinstance(other, type(self)) and addressof(self) == addressof(other) 
115   
116  t_espeak_callback=CFUNCTYPE(c_int,POINTER(c_short),c_int,POINTER(espeak_EVENT)) 
117   
118  @t_espeak_callback 
119 -def callback(wav,numsamples,event): 
120          try: 
121                  global player, isSpeaking, lastIndex 
122                  if not isSpeaking: 
123                          return 1 
124                  for e in event: 
125                          if e.type==espeakEVENT_MARK: 
126                                  lastIndex=int(e.id.name) 
127                          elif e.type==espeakEVENT_LIST_TERMINATED: 
128                                  break 
129                  if not wav: 
130                          player.idle() 
131                          isSpeaking = False 
132                          return 0 
133                  if numsamples > 0: 
134                          try: 
135                                  player.feed(string_at(wav, numsamples * sizeof(c_short))) 
136                          except: 
137                                  log.debugWarning("Error feeding audio to nvWave",exc_info=True) 
138                  return 0 
139          except: 
140                  log.error("callback", exc_info=True) 
141   
142 -class BgThread(threading.Thread): 
143 -        def __init__(self): 
144                  threading.Thread.__init__(self) 
145                  self.setDaemon(True) 
146   
147 -        def run(self): 
148                  global isSpeaking 
149                  while True: 
150                          func, args, kwargs = bgQueue.get() 
151                          if not func: 
152                                  break 
153                          try: 
154                                  func(*args, **kwargs) 
155                          except: 
156                                  log.error("Error running function from queue", exc_info=True) 
157                          bgQueue.task_done() 
158   
159 -def _execWhenDone(func, *args, **kwargs): 
160          global bgQueue 
161          # This can't be a kwarg in the function definition because it will consume the first non-keywor dargument which is meant for func. 
162          mustBeAsync = kwargs.pop("mustBeAsync", False) 
163          if mustBeAsync or bgQueue.unfinished_tasks != 0: 
164                  # Either this operation must be asynchronous or There is still an operation in progress. 
165                  # Therefore, run this asynchronously in the background thread. 
166                  bgQueue.put((func, args, kwargs)) 
167          else: 
168                  func(*args, **kwargs) 
169   
170 -def _speak(text): 
171          global isSpeaking 
172          uniqueID=c_int() 
173          isSpeaking = True 
174          flags = espeakCHARS_WCHAR | espeakSSML  
175          return espeakDLL.espeak_Synth(text,0,0,0,0,flags,byref(uniqueID),0) 
176   
177 -def speak(text): 
178          global bgQueue 
179          _execWhenDone(_speak, text, mustBeAsync=True) 
180   
181 -def stop(): 
182          global isSpeaking, bgQueue, lastIndex 
183          # Kill all speech from now. 
184          # We still want parameter changes to occur, so requeue them. 
185          params = [] 
186          try: 
187                  while True: 
188                          item = bgQueue.get_nowait() 
189                          if item[0] != _speak: 
190                                  params.append(item) 
191                          bgQueue.task_done() 
192          except Queue.Empty: 
193                  # Let the exception break us out of this loop, as queue.empty() is not reliable anyway. 
194                  pass 
195          for item in params: 
196                  bgQueue.put(item) 
197          isSpeaking = False 
198          player.stop() 
199          lastIndex=None 
200   
201 -def pause(switch): 
202          global player 
203          player.pause(switch) 
204   
205 -def setParameter(param,value,relative): 
206          _execWhenDone(espeakDLL.espeak_SetParameter,param,value,relative) 
207   
208 -def getParameter(param,current): 
209          return espeakDLL.espeak_GetParameter(param,current) 
210   
211 -def getVoiceList(): 
212          voices=espeakDLL.espeak_ListVoices(None) 
213          voiceList=[] 
214          for voice in voices: 
215                  if not voice: break 
216                  voiceList.append(voice.contents) 
217          return voiceList 
218   
219 -def getCurrentVoice(): 
220          voice =  espeakDLL.espeak_GetCurrentVoice() 
221          if voice: 
222                  return voice.contents 
223          else: 
224                  return None 
225   
226 -def setVoice(voice): 
227          # For some weird reason, espeak_EspeakSetVoiceByProperties throws an integer divide by zero exception. 
228          setVoiceByName(voice.identifier) 
229   
230 -def setVoiceByName(name): 
231          _execWhenDone(espeakDLL.espeak_SetVoiceByName,name) 
232   
233 -def _setVoiceAndVariant(voice=None, variant=None): 
234          res = getCurrentVoice().identifier.split("+") 
235          if not voice: 
236                  voice = res[0] 
237          if not variant: 
238                  if len(res) == 2: 
239                          variant = res[1] 
240                  else: 
241                          variant = "none" 
242          if variant == "none": 
243                  espeakDLL.espeak_SetVoiceByName(voice) 
244          else: 
245                  try: 
246                          espeakDLL.espeak_SetVoiceByName("%s+%s" % (voice, variant)) 
247                  except: 
248                          espeakDLL.espeak_SetVoiceByName(voice) 
249   
250 -def setVoiceAndVariant(voice=None, variant=None): 
251          _execWhenDone(_setVoiceAndVariant, voice=voice, variant=variant) 
252   
253 -def _setVoiceByLanguage(lang): 
254          v=espeak_VOICE() 
255          lang=lang.replace('_','-') 
256          v.languages=lang 
257          try: 
258                  espeakDLL.espeak_SetVoiceByProperties(byref(v)) 
259          except: 
260                  v.languages="en" 
261                  espeakDLL.espeak_SetVoiceByProperties(byref(v)) 
262   
263 -def setVoiceByLanguage(lang): 
264          _execWhenDone(_setVoiceByLanguage, lang) 
265   
266 -def espeak_errcheck(res, func, args): 
267          if res != EE_OK: 
268                  raise RuntimeError("%s: code %d" % (func.__name__, res)) 
269          return res 
270   
271 -def initialize(): 
272          global espeakDLL, bgThread, bgQueue, player 
273          espeakDLL=cdll.LoadLibrary(r"synthDrivers\espeak.dll") 
274          espeakDLL.espeak_Info.restype=c_char_p 
275          espeakDLL.espeak_Synth.errcheck=espeak_errcheck 
276          espeakDLL.espeak_SetVoiceByName.errcheck=espeak_errcheck 
277          espeakDLL.espeak_SetVoiceByProperties.errcheck=espeak_errcheck 
278          espeakDLL.espeak_SetParameter.errcheck=espeak_errcheck 
279          espeakDLL.espeak_Terminate.errcheck=espeak_errcheck 
280          espeakDLL.espeak_ListVoices.restype=POINTER(POINTER(espeak_VOICE)) 
281          espeakDLL.espeak_GetCurrentVoice.restype=POINTER(espeak_VOICE) 
282          espeakDLL.espeak_SetVoiceByName.argtypes=(c_char_p,) 
283          sampleRate=espeakDLL.espeak_Initialize(AUDIO_OUTPUT_SYNCHRONOUS,300,"synthDrivers",0) 
284          if sampleRate<0: 
285                  raise OSError("espeak_Initialize %d"%sampleRate) 
286          player = nvwave.WavePlayer(channels=1, samplesPerSec=sampleRate, bitsPerSample=16, outputDevice=config.conf["speech"]["outputDevice"]) 
287          espeakDLL.espeak_SetSynthCallback(callback) 
288          bgQueue = Queue.Queue() 
289          bgThread=BgThread() 
290          bgThread.start() 
291   
292 -def terminate(): 
293          global bgThread, bgQueue, player, espeakDLL  
294          stop() 
295          bgQueue.put((None, None, None)) 
296          bgThread.join() 
297          espeakDLL.espeak_Terminate() 
298          bgThread=None 
299          bgQueue=None 
300          player.close() 
301          player=None 
302          espeakDLL=None 
303   
304 -def info(): 
305          return espeakDLL.espeak_Info() 
306   
307 -def getVariantDict(): 
308          dir='synthDrivers\\espeak-data\\voices\\!v' 
309          variantDict={"none":_("none")} 
310          for fileName in os.listdir(dir): 
311                  if os.path.isfile("%s\\%s"%(dir,fileName)): 
312                          file=codecs.open("%s\\%s"%(dir,fileName)) 
313                          for line in file: 
314                                  if line.startswith('name '): 
315                                          temp=line.split(" ") 
316                                          if len(temp) ==2: 
317                                                  name=temp[1].rstrip() 
318                                                  break 
319                                  name=None 
320                          file.close() 
321                  if name is not None: 
322                          variantDict[fileName]=name 
323          return variantDict 
324