fixed most of the audio stuttering

README.md

@@ -1,25 +1,29 @@
-## Installation
+# M2 Projet Thématique - Diffusion Radio Proche-en-Proche
+
+Un projet visant à créer un réseau de machine capable de diffuser une source 
+audio à jouer de manière synchronisé.
+
+Les communications du réseau doivent être chiffré et il ne doit pas être possible
+d'inséré une machine inconnu pour pertuber le réseau.
+
+## Usage
 
 Debian
 ```bash
-# Pre-requires
-sudo apt install chrony
+# dependencies
+sudo apt install ffmpeg 
 
-# Download
-sudo apt install -y git
+# download the project
 git clone https://git.faraphel.fr/study-faraphel/M2-PT-DRP
-cd M2-PT-DRP
+cd ./M2-PT-DRP/
 
-# Dependencies
-sudo apt install -y libmpg123-dev libssl-dev portaudio19-dev
+# create a virtual environment
+python3 -m venv ./.venv/
+source ./.venv/bin/activate
 
-# Compile
-sudo apt install -y build-essential cmake ninja-build pkg-config
-cmake -S . -B build -G Ninja
-cmake --build build 
-cd build
+# install python packages
+pip3 install -r ./requirements.txt
 
-# Run
-sudo ./M2-PT-DRP --host ff02::1 --ipv6
+# run the application
+python3 -m source
 ```
-

source/behaviors/roles/MasterRole.py

@@ -3,6 +3,7 @@ from datetime import datetime, timedelta
 
 import pause
 import pydub
+from pydub.utils import make_chunks
 
 from source.behaviors.roles import base
 from source.managers import Manager
@@ -25,27 +26,24 @@ class MasterRole(base.BaseRole):
         self.manager.communication.secret_key = os.urandom(32)
 
         # prepare the audio file that will be streamed
+        # TODO(Faraphel): use another audio source
         self.audio = pydub.AudioSegment.from_file("../assets/Caravan Palace - Wonderland.mp3")
         self.play_time = datetime.now()
 
+        # calculate the number of bytes per milliseconds in the audio
+        bytes_per_ms = self.audio.frame_rate * self.audio.sample_width * self.audio.channels / 1000
+        # calculate the required chunk duration to reach that size
+        self.chunk_duration = timedelta(milliseconds=self.TARGET_SIZE / bytes_per_ms)
+
+        # split the audio into chunks
+        self.chunks = make_chunks(self.audio, self.chunk_duration.total_seconds() * 1000)
+
 
     def handle(self):
         # TODO(Faraphel): check if another server is emitting sound in the network. Return to undefined if yes
 
-        # calculate the number of bytes per milliseconds in the audio
-        bytes_per_ms = self.audio.frame_rate * self.audio.sample_width * self.audio.channels / 1000
-        # calculate the required chunk duration to reach that size
-        chunk_duration = timedelta(milliseconds=self.TARGET_SIZE / bytes_per_ms)
-
-        # calculate the audio time
-        chunk_start_time = datetime.now() - self.play_time
-        chunk_end_time = chunk_start_time + chunk_duration
-
-        # get the music for that period
-        chunk = self.audio[
-            chunk_start_time.total_seconds() * 1000 :
-            chunk_end_time.total_seconds() * 1000
-        ]
+        # get the next chunk
+        chunk = self.chunks.pop(0)
 
         # broadcast it in the network
         audio_packet = AudioPacket(
@@ -59,4 +57,4 @@ class MasterRole(base.BaseRole):
 
         # wait for the audio to play
         # TODO(Faraphel): should adapt to the compute time above
-        pause.until(self.play_time + chunk_end_time)
+        pause.until(datetime.now() + self.chunk_duration)

source/managers/AudioManager.py

@@ -1,4 +1,5 @@
 import threading
+import typing
 from datetime import datetime
 
 import numpy
@@ -12,6 +13,8 @@ from source.utils.audio.audio import sample_width_to_type
 
 class AudioManager:
     def __init__(self, manager: "managers.Manager"):
+        self.stream: typing.Optional[sounddevice.OutputStream] = None
+
         # buffer containing the set of audio chunk to play. Sort them by their time to play
         self.buffer = sortedcontainers.SortedList(key=lambda audio: audio.time)
 
@@ -31,13 +34,46 @@ class AudioManager:
             # trigger the new audio event
             self.new_audio_event.set()
 
+    def play_audio(self, audio: packets.AudioPacket) -> None:
+        # create a numpy array for our sample
+        sample = numpy.frombuffer(audio.data, dtype=sample_width_to_type(audio.sample_width))
+        # reshape it to have a sub-array for each channels
+        sample = sample.reshape((-1, audio.channels))
+        # normalize the sample to be between -1 and 1
+        sample = sample / (2 ** (audio.sample_width * 8 - 1))
+        # use float32 for the audio library
+        sample = sample.astype(numpy.float32)
+
+        # wait for the audio given time
+        pause.until(audio.time)
+
+        # update the stream if the audio use different settings
+        if (
+            self.stream is None or
+            self.stream.samplerate != audio.sample_rate or
+            self.stream.channels != audio.channels
+        ):
+            self.stream = sounddevice.OutputStream(
+                samplerate=audio.sample_rate,
+                channels=audio.channels,
+            )
+
+        # play
+        self.stream.start()
+
+        # write the audio to the stream
+        self.stream.write(sample)
+
     def handle(self) -> None:
         """
         Play the audio chunk in the buffer at the given time
         """
 
         # wait for a new audio packet
-        self.new_audio_event.wait()
+        # TODO(Faraphel): use self.lock ? seem to softlock the application
+        if len(self.buffer) == 0:
+            self.new_audio_event.clear()
+            self.new_audio_event.wait()
 
         # get the most recent audio packet to play
         audio: packets.AudioPacket = self.buffer.pop(0)
@@ -46,18 +82,9 @@ class AudioManager:
         if audio.time < datetime.now():
             return
 
-        # create a numpy array for our sample
-        sample = numpy.frombuffer(audio.data, dtype=sample_width_to_type(numpy.int16))
-        # reshape it to have a sub-array for each channels
-        sample = sample.reshape((-1, audio.channels))
-        # normalize the sample to be between -1 and 1
-        sample = sample / (2 ** (audio.sample_width * 8 - 1))
+        # play the audio packet
+        self.play_audio(audio)
 
-        # wait for the audio given time
-        pause.until(audio.time)
-
-        # play the audio
-        sounddevice.play(sample, audio.sample_rate)
 
     def loop(self) -> None:
         """