Improving Windows playback, sounddevice is default

Author: dschreij

mediadecoder/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.1.2"
+__version__ = "0.1.3"
 __author__ = "Daniel Schreij"
 __license__ = "MIT"
 

mediadecoder/decoder.py

@@ -415,41 +415,44 @@ def __audiorender_thread(self):
 
 		while self.status in [PLAYING,PAUSED]:
 			# Retrieve audiochunk
-			if self.status == PLAYING and new_audioframe is None:
-				# Get a new frame from the audiostream, skip to the next one
-				# if the current one gives a problem
-				try:
-					start = self.audio_times.pop(0)
-					stop = self.audio_times[0]
-				except IndexError:
-					logger.debug("Audio times could not be obtained")
-					time.sleep(0.02)
-					continue
-
-				# Get the frame numbers to extract from the audio stream.
-				chunk = (1.0/self.audioformat['fps'])*np.arange(start, stop)
-
-				try:
-					# Extract the frames from the audio stream. Does not always,
-					# succeed (e.g. with bad streams missing frames), so make
-					# sure this doesn't crash the whole program.
-					new_audioframe = self.clip.audio.to_soundarray(
-						tt = chunk,
-						buffersize = self.frame_interval*self.clip.audio.fps,
-						quantize=True
-					)
-				except OSError as e:
-					logger.warning("Sound decoding error: {}".format(e))
-					new_audioframe = None
-			# Put audioframe in buffer/queue for soundrenderer to pick up. If
-			# the queue is full, try again after a timeout (this allows to check
-			# if the status is still PLAYING after a pause.)
-			if not new_audioframe is None:
-				try:
-					self.audioqueue.put(new_audioframe, timeout=.05)
-					new_audioframe = None
-				except Full:
-					pass
+			if self.status == PLAYING:
+				if new_audioframe is None:
+					# Get a new frame from the audiostream, skip to the next one
+					# if the current one gives a problem
+					try:
+						start = self.audio_times.pop(0)
+						stop = self.audio_times[0]
+					except IndexError:
+						logger.debug("Audio times could not be obtained")
+						time.sleep(0.02)
+						continue
+
+					# Get the frame numbers to extract from the audio stream.
+					chunk = (1.0/self.audioformat['fps'])*np.arange(start, stop)
+
+					try:
+						# Extract the frames from the audio stream. Does not always,
+						# succeed (e.g. with bad streams missing frames), so make
+						# sure this doesn't crash the whole program.
+						new_audioframe = self.clip.audio.to_soundarray(
+							tt = chunk,
+							buffersize = self.frame_interval*self.clip.audio.fps,
+							quantize=True
+						)
+					except OSError as e:
+						logger.warning("Sound decoding error: {}".format(e))
+						new_audioframe = None
+				# Put audioframe in buffer/queue for soundrenderer to pick up. If
+				# the queue is full, try again after a timeout (this allows to check
+				# if the status is still PLAYING after a pause.)
+				if not new_audioframe is None:
+					try:
+						self.audioqueue.put(new_audioframe, timeout=.05)
+						new_audioframe = None
+					except Full:
+						pass
+			
+			time.sleep(0.005)
 
 		logger.debug("Stopped audio rendering thread.")
 

mediadecoder/soundrenderers/pyaudiorenderer.py

@@ -50,8 +50,7 @@ def get_frame(self, in_data, frame_count, time_info, status):
 				frame = self.queue.get(False, timeout=queue_timeout)
 				return (frame, pyaudio.paContinue)
 			except Empty:
-				time.sleep(queue_timeout)
-				continue
+				pass
 		return (None, pyaudio.paComplete)
 
 	def start(self):

mediadecoder/soundrenderers/pyaudiorenderer2.py

@@ -0,0 +1,69 @@
+import time
+import pyaudio
+import threading
+
+try:
+	# Python 3
+	from queue import Queue, Empty
+except:
+	# Python 2
+	from Queue import Queue, Empty
+
+from mediadecoder.soundrenderers._base import SoundRenderer
+
+queue_timeout=0.01
+
+class SoundrendererPyAudio(threading.Thread, SoundRenderer):
+	""" Uses pyaudio to play sound """
+	def __init__(self, audioformat, queue=None):
+		"""Constructor.
+		Creates a pyaudio sound renderer.
+		
+		Parameters
+		----------
+		audioformat : dict
+			A dictionary containing the properties of the audiostream
+		queue : Queue.queue
+			A queue object which serves as a buffer on which the individual
+			audio frames are placed by the decoder.
+		"""
+		if pyaudio is None:
+			raise RuntimeError("Pyaudio sound renderer is not available")
+
+		super(SoundrendererPyAudio, self).__init__()
+
+		if not queue is None:
+			self.queue = queue
+
+		self.pa = pyaudio.PyAudio()
+		self.stream = self.pa.open(
+			channels  	= audioformat["nchannels"],
+			rate 		= audioformat["fps"],
+			# frames_per_buffer = audioformat['buffersize'],
+			format 	= pyaudio.get_format_from_width(audioformat["nbytes"]),
+			output 	= True,
+		)
+
+	def run(self):
+		""" Initializes the stream. """
+		if not hasattr(self, 'queue'):
+			raise RuntimeError("Audio queue is not intialized.")
+
+		self.stream.start_stream()
+		self.keep_listening = True	
+		
+		while self.keep_listening:
+			try:
+				frame = self.queue.get(False, timeout=queue_timeout)
+				self.stream.write(frame)
+			except Empty:
+				continue
+			time.sleep(0.01)
+
+		self.stream.stop_stream()
+		self.stream.close()
+		self.pa.terminate()
+
+	def close_stream(self):
+		""" Closes the stream. Performs cleanup. """
+		self.keep_listening = False

mediadecoder/soundrenderers/pygamerenderer.py

@@ -1,3 +1,4 @@
+import time
 import threading
 import pygame
 
@@ -68,6 +69,7 @@ def run(self):
 				if not channel.get_queue():	
 					channel.queue(chunk)
 					chunk = None
+			time.sleep(0.005)
 
 		if not channel is None:
 			channel.stop()

play.py

@@ -134,10 +134,10 @@ def load_media(self, vidSource):
 				from mediadecoder.soundrenderers import SoundrendererPygame
 				self.audio = SoundrendererPygame(self.decoder.audioformat)
 			elif self.soundrenderer == "pyaudio":
-				from mediadecoder.soundrenderers import SoundrendererPyAudio
+				from mediadecoder.soundrenderers.pyaudiorenderer import SoundrendererPyAudio
 				self.audio = SoundrendererPyAudio(self.decoder.audioformat)
 			elif self.soundrenderer == "sounddevice":
-				from mediadecoder.soundrenderers import SoundrendererSounddevice
+				from mediadecoder.soundrenderers.sounddevicerenderer import SoundrendererSounddevice
 				self.audio = SoundrendererSounddevice(self.decoder.audioformat)
 			self.decoder.set_audiorenderer(self.audio)
 
@@ -293,8 +293,8 @@ def pause(self):
 	parser.add_argument("-l", "--loop", help="loop the video", 
 		action="store_true", default=False)
 	parser.add_argument("-s", "--soundrenderer", help="the backend that should "
-		" render the sound (default: pyaudio)", choices=["pygame","pyaudio",
-		"sounddevice"], default="pyaudio")
+		" render the sound (default: sounddevice)", choices=["pygame","pyaudio",
+		"sounddevice"], default="sounddevice")
 	parser.add_argument("-r", "--resolution", help="The resolution of the video."
 		"\nSpecify as <width>x<height> (default: 1024x768)", default="1024x768")