194 lines
7.0 KiB
Python
194 lines
7.0 KiB
Python
import audioop
|
|
import math
|
|
import pyaudio
|
|
import asyncio
|
|
from internal.logger import create_logger
|
|
from discord import AudioSource
|
|
|
|
LOGGER = create_logger(__name__)
|
|
|
|
# Constants for audio processing
|
|
SAMPLES_PER_FRAME = 960
|
|
CHANNELS = 2
|
|
SAMPLE_RATE = 48000
|
|
FRAME_SIZE = SAMPLES_PER_FRAME * CHANNELS * 2 # 16-bit PCM
|
|
SILENT_FRAME = b'\x00' * FRAME_SIZE
|
|
|
|
|
|
class NoiseGateSource(AudioSource):
|
|
def __init__(self, audio_stream, threshold: int):
|
|
self.audio_stream = audio_stream
|
|
self.threshold = threshold
|
|
self.ng_fadeout_count = 0
|
|
self.NG_FADEOUT_FRAMES = 12 # 240ms fadeout time
|
|
|
|
def read(self) -> bytes:
|
|
"""
|
|
Reads data from the audio stream, applies the noise gate,
|
|
and returns a 20ms audio frame.
|
|
"""
|
|
try:
|
|
# Read a frame's worth of data from the input stream.
|
|
pcm_data = self.audio_stream.read(SAMPLES_PER_FRAME, exception_on_overflow=False)
|
|
|
|
# Ensure we have a full frame of data.
|
|
if len(pcm_data) != FRAME_SIZE:
|
|
return
|
|
|
|
# Calculate volume to check against the threshold.
|
|
rms = audioop.rms(pcm_data, 2)
|
|
if rms == 0:
|
|
# If there's no volume, check if we're in the fadeout period.
|
|
if self.ng_fadeout_count > 0:
|
|
self.ng_fadeout_count -= 1
|
|
return pcm_data # Return the (silent) data to complete the fade
|
|
return
|
|
|
|
db = 20 * math.log10(rms)
|
|
|
|
# If volume is above the threshold, send the audio and reset fadeout.
|
|
if db >= self.threshold:
|
|
self.ng_fadeout_count = self.NG_FADEOUT_FRAMES
|
|
return pcm_data
|
|
|
|
# If below threshold but still in the fadeout period, send the audio.
|
|
if self.ng_fadeout_count > 0:
|
|
self.ng_fadeout_count -= 1
|
|
return pcm_data
|
|
|
|
# Otherwise, the gate is closed.
|
|
return
|
|
|
|
except Exception as e:
|
|
LOGGER.error(f"Error in NoiseGateSource.read: {e}", exc_info=True)
|
|
return SILENT_FRAME
|
|
|
|
def cleanup(self) -> None:
|
|
"""Called when the player stops."""
|
|
if self.audio_stream:
|
|
self.audio_stream.stop_stream()
|
|
self.audio_stream.close()
|
|
LOGGER.info("Audio stream cleaned up.")
|
|
|
|
class AudioStreamManager:
|
|
"""Manages the PyAudio instance and input stream."""
|
|
def __init__(self, input_device_index: int):
|
|
self.pa = pyaudio.PyAudio()
|
|
self.stream = self.pa.open(
|
|
format=pyaudio.paInt16,
|
|
channels=CHANNELS,
|
|
rate=SAMPLE_RATE,
|
|
input=True,
|
|
frames_per_buffer=SAMPLES_PER_FRAME,
|
|
input_device_index=input_device_index
|
|
)
|
|
self.stream.start_stream()
|
|
LOGGER.info(f"Audio stream started on device {input_device_index}")
|
|
|
|
def get_stream(self):
|
|
return self.stream
|
|
|
|
def terminate(self):
|
|
if self.stream and self.stream.is_active():
|
|
self.stream.stop_stream()
|
|
self.stream.close()
|
|
LOGGER.debug("[ReopenStream.close_if_open]:\t Stream was open; It was closed.")
|
|
|
|
def list_devices(self, _display_input_devices: bool = True, _display_output_devices: bool = True):
|
|
LOGGER.info('Getting a list of the devices connected')
|
|
info = self.paInstance.get_host_api_info_by_index(0)
|
|
numdevices = info.get('deviceCount')
|
|
|
|
devices = {'Input': {}, 'Output': {}}
|
|
for i in range(0, numdevices):
|
|
device_info = self.paInstance.get_device_info_by_host_api_device_index(0, i)
|
|
if (device_info.get('maxInputChannels')) > 0:
|
|
input_device = device_info.get('name')
|
|
devices['Input'][i] = input_device
|
|
if _display_input_devices:
|
|
LOGGER.debug(f"Input Device id {i} - {input_device}")
|
|
|
|
if (device_info.get('maxOutputChannels')) > 0:
|
|
output_device = device_info.get('name')
|
|
devices['Output'][i] = output_device
|
|
if _display_output_devices:
|
|
LOGGER.debug(f"Output Device id {i} - {output_device}")
|
|
return devices
|
|
|
|
# noinspection PyUnresolvedReferences
|
|
class NoiseGate(AudioStream):
|
|
def __init__(self, _voice_connection, _noise_gate_threshold: int, **kwargs):
|
|
super(NoiseGate, self).__init__(_init_on_startup=True, **kwargs)
|
|
self.voice_connection = _voice_connection
|
|
self.THRESHOLD = _noise_gate_threshold
|
|
self.NGStream = NoiseGateStream(self)
|
|
|
|
def run(self) -> None:
|
|
LOGGER.debug("Starting stream")
|
|
self.stream.start_stream()
|
|
self.core()
|
|
|
|
def core(self):
|
|
if self.voice_connection.is_connected() and not self.voice_connection.is_playing():
|
|
LOGGER.debug("Playing stream to discord")
|
|
self.voice_connection.play(self.NGStream)
|
|
|
|
async def close(self):
|
|
LOGGER.debug("Closing NoiseGate resources...")
|
|
if self.voice_connection and self.voice_connection.is_connected():
|
|
self.voice_connection.stop()
|
|
|
|
self.close_if_open()
|
|
|
|
if self.paInstance:
|
|
self.paInstance.terminate()
|
|
|
|
LOGGER.debug("NoiseGate resources closed.")
|
|
|
|
# noinspection PyUnresolvedReferences
|
|
class NoiseGateStream(discord.AudioSource):
|
|
def __init__(self, noise_gate_instance: NoiseGate):
|
|
super(NoiseGateStream, self).__init__()
|
|
self.noise_gate = noise_gate_instance
|
|
self.NG_fadeout = 12
|
|
self.NG_fadeout_count = 0
|
|
self.process_set_count = 0
|
|
|
|
def read(self):
|
|
try:
|
|
if not self.noise_gate.voice_connection.is_connected():
|
|
return SILENT_FRAME
|
|
|
|
curr_buffer = self.noise_gate.stream.read(960, exception_on_overflow=False)
|
|
|
|
if len(curr_buffer) != DISCORD_FRAME_SIZE:
|
|
return SILENT_FRAME
|
|
|
|
buffer_rms = audioop.rms(curr_buffer, 2)
|
|
|
|
if buffer_rms > 0:
|
|
buffer_decibel = 20 * math.log10(buffer_rms)
|
|
|
|
if self.process_set_count % 10 == 0:
|
|
log_msg = f"[{'Open' if buffer_decibel >= self.noise_gate.THRESHOLD else 'Closed'}]"
|
|
LOGGER.debug(f"[NoiseGate {log_msg}] {buffer_decibel:.2f} dB")
|
|
|
|
if buffer_decibel >= self.noise_gate.THRESHOLD:
|
|
self.NG_fadeout_count = self.NG_fadeout
|
|
self.process_set_count += 1
|
|
return bytes(curr_buffer)
|
|
|
|
elif self.NG_fadeout_count > 0:
|
|
self.NG_fadeout_count -= 1
|
|
self.process_set_count += 1
|
|
return bytes(curr_buffer)
|
|
|
|
return SILENT_FRAME
|
|
|
|
except IOError as e:
|
|
LOGGER.error(f"PyAudio IOError in read(): {e}")
|
|
return SILENT_FRAME
|
|
except Exception as e:
|
|
LOGGER.error(f"Unhandled exception in NoiseGateStream.read: {e}", exc_info=True)
|
|
return SILENT_FRAME
|