乐趣区

关于python:视频静音的一种检测方案

利用 AudioSegment 库抽出音轨,肯定时间段距离采样音频的 dBFS,如果小于阈值(个别认为小于 -50 为静音),当然还存在负无穷的景象须要判断。

def audio_silence_check_service(video_location):
    download_video_to_local(location=video_location, video_filename=video_filename)
    extract_audio_result = extract_audio_from_video(video_filename, audio_filename)
    is_exist_silence = False
    silence_desc = ''
    silence_detect_result = []
    if extract_audio_result:
        silence_detect_result = silence_detect(audio_filename)
        for silence_chunk in silence_detect_result.get('silence_result'):
            if int(silence_chunk[1]) - int(silence_chunk[0]) > int(int(silence_detect_result.get('audio_duration')) / 3):
                silence_desc += f'{silence_chunk[0]}~{silence_chunk[1]},'
    if silence_desc:
        silence_desc += 'ms 存在静音'
        is_exist_silence = True
    else:
        is_exist_silence = True
        silence_desc = f'{qipuid} 未获取到音轨'
        logger_error(silence_desc)
        return is_exist_silence, silence_desc, silence_detect_result
        
def silence_detect(audio_file, silence_threshold=-50, chunk_size=10):
    sound = AudioSegment.from_file(audio_file, format="wav")
    assert chunk_size > 0
    silence_chunks = list()
    silence_chunks_start_ms = -1
    sound_duration = len(sound)
    for current_ms in range(0, sound_duration, chunk_size):
        current_ms_chunks_dBFS = sound[current_ms: current_ms + chunk_size].dBFS
        if current_ms_chunks_dBFS < silence_threshold or current_ms_chunks_dBFS == -inf:
            if silence_chunks_start_ms == -1:
                silence_chunks_start_ms = current_ms
            else:
                if current_ms <= sound_duration <= current_ms + chunk_size:
                    silence_chunks.append([silence_chunks_start_ms, sound_duration])
                    silence_chunks_start_ms = -1
        else:
            if silence_chunks_start_ms != -1:
                silence_chunks.append([silence_chunks_start_ms, current_ms])
                silence_chunks_start_ms = -1

    return {'audio_duration': sound_duration, 'silence_result': silence_chunks}
    
def download_video_to_local(location, video_filename):
    res = get(location)
    with open(video_filename, r'wb') as f:
        f.write(res.content) 
退出移动版