Deepfake-Audio / Source Code /encoder_preprocess.py
ameythakur's picture
Deepfake-Audio
1d8403e verified
# ==================================================================================================
# DEEPFAKE AUDIO - encoder_preprocess.py (Acoustic Feature Extraction)
# ==================================================================================================
#
# πŸ“ DESCRIPTION
# This script serves as the primary data ingestion layer for the Speaker Encoder. It
# processes raw waveforms from massive speech datasets (LibriSpeech, VoxCeleb) and extracts
# high-dimensional Mel-Spectrogram features. These features are essentially "voice prints"
# that allow the encoder to learn speaker-independent embedding representations.
#
# πŸ‘€ AUTHORS
# - Amey Thakur (https://github.com/Amey-Thakur)
# - Mega Satish (https://github.com/msatmod)
#
# 🀝🏻 CREDITS
# Original Real-Time Voice Cloning methodology by CorentinJ
# Repository: https://github.com/CorentinJ/Real-Time-Voice-Cloning
#
# πŸ”— PROJECT LINKS
# Repository: https://github.com/Amey-Thakur/DEEPFAKE-AUDIO
# Video Demo: https://youtu.be/i3wnBcbHDbs
# Research: https://github.com/Amey-Thakur/DEEPFAKE-AUDIO/blob/main/DEEPFAKE-AUDIO.ipynb
#
# πŸ“œ LICENSE
# Released under the MIT License
# Release Date: 2021-02-06
# ==================================================================================================
from encoder.preprocess import preprocess_librispeech, preprocess_voxceleb1, preprocess_voxceleb2
from utils.argutils import print_args
from pathlib import Path
import argparse
if __name__ == "__main__":
# --- INTERFACE CONFIGURATION ---
class MyFormatter(argparse.ArgumentDefaultsHelpFormatter, argparse.RawDescriptionHelpFormatter):
"""Custom formatter to preserve formatting in the help description."""
pass
parser = argparse.ArgumentParser(
description="Acoustic Pre-processor: Normalizes raw datasets into mel-spectrogram arrays.\n"
"Required datasets: LibriSpeech, VoxCeleb1, or VoxCeleb2.",
formatter_class=MyFormatter
)
# --- PATH DEFINITIONS ---
parser.add_argument("datasets_root", type=Path,
help="Root directory where raw datasets are extracted.")
parser.add_argument("-o", "--out_dir", type=Path, default=argparse.SUPPRESS,
help="Destination for processed neural features. Defaults to <datasets_root>/SV2TTS/encoder/")
# --- PROCESSING PARAMETERS ---
parser.add_argument("-d", "--datasets", type=str,
default="librispeech_other,voxceleb1,voxceleb2",
help="Comma-separated identifiers of datasets to include in the pipeline.")
parser.add_argument("-s", "--skip_existing", action="store_true",
help="Optimize by skipping files that have already been materialized on disk.")
parser.add_argument("--no_trim", action="store_true",
help="Inhibit silent period removal (Voice Activity Detection). Not recommended for high-fidelity training.")
args = parser.parse_args()
# --- HARDWARE & VAD VALIDATION ---
# We verify the presence of 'webrtcvad' as it is critical for ensuring non-silent training samples.
if not hasattr(args, "no_trim") or not args.no_trim:
try:
import webrtcvad
except:
print("⚠️ Scholarly Warning: 'webrtcvad' not detected. This is required for speech silence removal.")
raise ModuleNotFoundError("Please install 'webrtcvad' or use --no_trim for a degraded run.")
# --- ARCHITECTURAL ORCHESTRATION ---
args.datasets = args.datasets.split(",")
if not hasattr(args, "out_dir"):
args.out_dir = args.datasets_root.joinpath("SV2TTS", "encoder")
assert args.datasets_root.exists(), "Fatal: datasets_root not found. 🀝🏻 Ensure pathing."
args.out_dir.mkdir(exist_ok=True, parents=True)
# --- EXECUTION ---
print_args(args, parser)
print("🀝🏻 Scholarly Partnership: Amey Thakur & Mega Satish")
preprocess_func = {
"librispeech_other": preprocess_librispeech,
"voxceleb1": preprocess_voxceleb1,
"voxceleb2": preprocess_voxceleb2,
}
args_dict = vars(args)
datasets_to_process = args_dict.pop("datasets")
for dataset in datasets_to_process:
print("\nπŸš€ Initiating Neural Feature Extraction for: %s" % dataset)
preprocess_func[dataset](**args_dict)