Python UDP audio streaming in real time

2 files changed, 89 insertions, 0 deletions

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..f7275bb
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+venv/
diff --git a/udp-audio.py b/udp-audio.py
new file mode 100644
index 0000000..d6d86a0
--- /dev/null
+++ b/udp-audio.py
@@ -0,0 +1,88 @@
+import socket
+import struct
+import pyaudio
+import time
+
+from threading import Thread
+import numpy as np
+import scipy.signal
+from matplotlib import pyplot as plt
+
+TGID_in_stream = True  # When set to True, we expect a 4 byte long int with the TGID prior to the audio in each packet
+TGID_to_play = 0  # When TGID_in_stream is set to True, we'll only play audio if the received TGID matches this value
+UDP_PORT = 9123  # UDP port to listen on
+AUDIO_OUTPUT_DEVICE_INDEX = 4  # Audio device to play received audio on
+
+# Set up a UDP server
+UDPSock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+# UDPSock.settimeout(1)
+
+listen_addr = ("", UDP_PORT)
+UDPSock.bind(listen_addr)
+
+p = pyaudio.PyAudio()
+
+for i in range(p.get_device_count()):
+    print(p.get_device_info_by_index(i))
+
+CHUNK=1024
+FORMAT = pyaudio.paInt16
+CHANNELS = 1
+IN_RATE = 8_000
+OUT_RATE = 48_000
+stream = p.open(
+    format=FORMAT,
+    channels=CHANNELS,
+    rate=OUT_RATE,
+    input=False,
+    output=True,
+    frames_per_buffer=CHUNK,
+    output_device_index=AUDIO_OUTPUT_DEVICE_INDEX,
+)
+
+FRAMES = []
+
+def receive_audio():
+    global FRAMES
+    global TGID_in_stream
+    global TGID_to_play
+    while True:
+        try:
+            data, addr = UDPSock.recvfrom((CHUNK * CHANNELS * 2) + (4 if TGID_in_stream else 0))
+            if TGID_in_stream:
+                tgid = int.from_bytes(data[0:4], "little")
+                if TGID_to_play == 0:
+                    TGID_to_play = [tgid]
+                if TGID_to_play == 0 or tgid in TGID_to_play:
+                    audio_data = data[4:]
+                    if len(audio_data) < 128:
+                        continue
+                    print(f"TGID: {tgid}, audio_data: {len(audio_data)}")
+                    signal = np.frombuffer(audio_data, dtype=np.int16)
+                    res = scipy.signal.resample(signal, (len(signal) * OUT_RATE) // IN_RATE).astype(np.int16)
+                    FRAMES.append(res)
+            else:
+                stream.write(data)
+        except socket.timeout:
+            pass
+
+def play_audio():
+    global FRAMES
+    while True:
+        while len(FRAMES) == 0:
+            time.sleep(1)
+        frames = FRAMES[:]
+        FRAMES.clear()
+        print(f"Playing {len(frames)} frames")
+        for frame in frames:
+            stream.write(frame.tobytes())
+
+
+stream_t = Thread(target=receive_audio)
+play_t = Thread(target=play_audio)
+stream_t.daemon = True
+play_t.daemon = True
+stream_t.start()
+play_t.start()
+stream_t.join()
+play_t.join()