data/Whisper-Sidecar-data-metadata/select_prompt_wav.py

import os
import glob 
import random 

import jsonlines

librispeech_dir = "./dataset/LibriSpeech/*"
enroll_path = "./prompt_enroll_audio/all"

if not os.path.exists(enroll_path):
    os.makedirs(enroll_path)

speaker_dirs = glob.glob(f"{librispeech_dir}/*")
new_files = []
for speaker_dir in speaker_dirs:
    if not "train-" in speaker_dir and not "test-" in speaker_dir and not "dev-" in speaker_dir:
        continue
    speaker_id = os.path.basename(speaker_dir)
    enroll_speaker_dir = os.path.join(enroll_path, speaker_id)
    if not os.path.exists(enroll_speaker_dir):
        os.makedirs(enroll_speaker_dir, exist_ok=True)
    # Check whether each speaker in librispeech_dir exists in enroll_path
    if len(glob.glob(f"{enroll_speaker_dir}/*.wav")) == 0:
        # Copy an audio file from librispeech_dir to enroll_speaker_dir
        flac_files = glob.glob(f"{speaker_dir}/*/*.flac")
        try:
            flac_files = random.sample(flac_files, 5)
        except:
            print()
            print(flac_files)
            print(speaker_dir)
        for flac_file in flac_files:
            new_flac_file = os.path.join(enroll_speaker_dir, os.path.basename(flac_file))
            # print(new_flac_file)
            os.system(f"cp {flac_file} {new_flac_file}")
            # Record the name of the flac_file
            new_files.append(flac_file)
    else:
        print(glob.glob(f"{enroll_speaker_dir}/*.wav"))
    
    # Check whether the voice naming of the speaker in enroll_path is {speaker_id}.wav
    enroll_wav_files = glob.glob(f"{enroll_speaker_dir}/*")
    for i, enroll_wav_file in enumerate(enroll_wav_files):
        wav_file = os.path.join(enroll_speaker_dir, speaker_id + f"_{i}.wav")
        os.system(f"ffmpeg -i {enroll_wav_file}  -t 3  {wav_file}")
        os.system(f"rm -rf {enroll_wav_file}")

# Record the original name of the enroll wav
with open(f"{enroll_path}/enrolled_wavs.txt", "w") as f:
    f.write("\n".join(new_files))