Concatena tutti gli audio di un dataset video in un unico file, aggiungendo anche il nome del file nel file audio (usando il comando "say" di macOS)

concat_audio_advanced.py

#!/usr/bin/env python3
import os
import subprocess
from tqdm import tqdm
from concurrent.futures import ThreadPoolExecutor

# === Config ===
VIDEO_EXT = ".mp4"
TEMP_SPEECH = "temp_speech"
TEMP_AUDIO = "temp_audio"
TEMP_LABELED = "temp_labeled"
OUTPUT_FILE = "output_labeled.m4a"

os.makedirs(TEMP_SPEECH, exist_ok=True)
os.makedirs(TEMP_AUDIO, exist_ok=True)
os.makedirs(TEMP_LABELED, exist_ok=True)

videos = sorted([f for f in os.listdir('.') if f.endswith(VIDEO_EXT)])

def process_video(f):
    base = os.path.splitext(f)[0]
    speech_path = os.path.join(TEMP_SPEECH, f"{base}.m4a")
    audio_path = os.path.join(TEMP_AUDIO, f"{base}.m4a")
    
    # Generate spoken filename
    aiff_tmp = os.path.join(TEMP_SPEECH, f"{base}.aiff")
    subprocess.run(['say', '-v', 'Samantha', base, '-o', aiff_tmp], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
    subprocess.run(['ffmpeg', '-y', '-i', aiff_tmp, '-ar', '44100', '-c:a', 'aac', '-b:a', '128k', '-ac', '2', speech_path],
                    stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
    os.remove(aiff_tmp)
        
    # Extract audio
    subprocess.run([
        'ffmpeg', '-y', '-i', f,
        '-vn', '-ac', '2', '-ar', '44100',
        '-c:a', 'aac', '-b:a', '128k',
        audio_path
    ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
        
# === Parallel Processing ===
with ThreadPoolExecutor() as executor:
    list(tqdm(executor.map(process_video, videos), total=len(videos), desc="Processing videos"))
    
# === Merge speech + audio ===
labeled_files = []
for f in tqdm(videos, desc="Merging labeled clips"):
    base = os.path.splitext(f)[0]
    speech = os.path.join(TEMP_SPEECH, f"{base}.m4a")
    audio = os.path.join(TEMP_AUDIO, f"{base}.m4a")
    labeled = os.path.join(TEMP_LABELED, f"{base}.m4a")
    
    subprocess.run([
        'ffmpeg', '-y',
        '-i', speech,
        '-i', audio,
        '-filter_complex', '[0:a][1:a]concat=n=2:v=0:a=1[outa]',
        '-map', '[outa]',
        '-c:a', 'aac', '-b:a', '128k', '-ac', '2', '-ar', '44100',
        labeled
    ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
    
    labeled_files.append(labeled)

# === Final Concatenation ===
with open("master_list.txt", "w") as master:
    for f in labeled_files:
        master.write(f"file '{os.path.abspath(f)}'\n")
        
# Try -c copy first, fallback to re-encoding if it fails
result = subprocess.run(['ffmpeg', '-y', '-f', 'concat', '-safe', '0', '-i', 'master_list.txt', '-c', 'copy', OUTPUT_FILE],
                        stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)

if result.returncode != 0:
    print("🔁 Re-encoding output due to stream incompatibility...")
    subprocess.run(['ffmpeg', '-y', '-f', 'concat', '-safe', '0', '-i', 'master_list.txt',
                    '-c:a', 'aac', '-b:a', '128k', '-ac', '2', OUTPUT_FILE],
                   stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
    
# Cleanup
os.remove("master_list.txt")
os.system(f"rm -rf {TEMP_AUDIO}")
os.system(f"rm -rf {TEMP_LABELED}")
os.system(f"rm -rf {TEMP_SPEECH}")

print(f"✅ Done! Final file: {OUTPUT_FILE}")