File size: 3,436 Bytes
b36d167 82f19d8 b36d167 82f19d8 b36d167 82f19d8 b36d167 08e515a b36d167 82f19d8 b36d167 82f19d8 b36d167 82f19d8 b36d167 82f19d8 b36d167 82f19d8 b36d167 08e515a b36d167 08e515a b36d167 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 |
import random
from pydub import AudioSegment # type: ignore
# from pydub.effects import normalize
# import numpy as np # type: ignore
def get_audio_volume_db(audio):
"""Estimate the volume in dBFS (decibels relative to full scale) using PyDub.""" # noqa
return audio.dBFS if audio.dBFS != float('-inf') else -50.0 # Default to -50 dB for silence # noqa
def adjust_volume(audio, volume_change_db):
"""Adjusts the volume of an AudioSegment."""
return audio + volume_change_db
# def compress_audio(audio):
# """Apply compression to normalize speech volume."""
# return normalize(audio)
def place_in_stereo(audio, pan_value):
"""Places audio in stereo field (-1.0 = full left, 1.0 = full right)."""
return audio.pan(pan_value)
def overlay_audio(speech_audio, noise_audio):
"""Overlays speech and noise using PyDub."""
return speech_audio.overlay(noise_audio)
def process_audio(speech_data, noise_data, speech_sr, noise_sr, alpha, beta):
"""
Process speech and noise audio data with quality preservation.
Args:
speech_data (numpy.ndarray): Speech audio data
noise_data (numpy.ndarray): Noise audio data
speech_sr (int): Speech sample rate
noise_sr (int): Noise sample rate
alpha (float): Speech volume adjustment
beta (float): Noise volume adjustment
"""
# Convert numpy arrays to AudioSegment
speech_audio = AudioSegment(
speech_data.tobytes(),
frame_rate=speech_sr,
sample_width=speech_data.dtype.itemsize,
channels=1
)
noise_audio = AudioSegment(
noise_data.tobytes(),
frame_rate=noise_sr,
sample_width=noise_data.dtype.itemsize,
channels=1
)
# Get speech duration
speech_duration = len(speech_audio) / 1000.0 # Convert ms to sec
# Modify crossfade duration based on audio length
crossfade_duration = min(5, len(speech_audio) // 4) # Use 5ms or 1/4 of audio length, whichever is smaller
# Cut noise segment with crossfade to avoid clicks
if len(noise_audio) / 1000.0 <= speech_duration:
trimmed_noise = noise_audio
else:
start_time = random.uniform(0, len(noise_audio) / 1000.0 - speech_duration) * 1000
trimmed_noise = noise_audio[start_time:start_time + (speech_duration * 1000)]
# Adjust crossfade duration for short clips
trimmed_noise = trimmed_noise.fade_in(crossfade_duration).fade_out(crossfade_duration)
# Match sample rates before mixing
trimmed_noise = trimmed_noise.set_frame_rate(speech_sr)
# Gradual volume adjustment with adjusted crossfade
adjusted_speech = speech_audio
if alpha != 0:
adjusted_speech = adjust_volume(speech_audio, alpha).fade_in(crossfade_duration).fade_out(crossfade_duration)
adjusted_noise = trimmed_noise
if beta != 0:
adjusted_noise = adjust_volume(trimmed_noise, beta).fade_in(crossfade_duration).fade_out(crossfade_duration)
# Overlay with crossfade to preserve quality
final_audio = adjusted_speech.overlay(adjusted_noise, gain_during_overlay=0)
return final_audio
# final_audio = process_audio("anushka.wav", "traffic.wav")
# # Single write operation at the end
# final_audio.export("output-traffic.wav", format="wav")
# print("Processing complete. Check output.wav!")
# -18, -20 for office
# -13 , -20 for market
# -18, -20 for traffic
|