从MP4文件中的语音提取文本，主要有2个核心逻辑：

1.从mp4文件中获取语音

2.语音转文本

从mp4文件中获取语音的python实现如下：

import moviepy.editor as mp
def convert_mp4_to_wav(input_file, output_file):
    video = mp.VideoFileClip(input_file)
    video.audio.write_audiofile(output_file)
if __name__ == '__main__':
    file_name="1.mp4"
    convert_mp4_to_wav(file_name,'1.wav')

语音转文本，这个技术也比较成熟了，我这边使用了stt中的models--Systran--faster-whisper-medium模型，准确率不错。

忽略stt具体使用，如果有需要，就下载代码。

这里说一下，怎么调用stt并输出文本，其代码如下：

def mp4totxt(mp4FileName):
    #添加日志
    logging.basicConfig(filename='mp4totxt.log', level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s') 
    
    logging.info("--开始----"+mp4FileName)
    wavFileName=mp4FileName.lower().replace('.mp4','.wav')
    #mp4转wav
    convert_mp4_to_wav(mp4FileName,wavFileName)
    url = "http://127.0.0.1:9971/api"
    # 请求参数 file:音视频文件，language：语言代码，model：模型，response_format:text|json|srt
    # 返回 code==0 成功，其他失败，msg==成功为ok，其他失败原因，data=识别后返回文字
    files = {"file": open(wavFileName, "rb")}
    data={"language":"zh","model":"medium","response_format":"json"}
    response = requests.request("POST", url, timeout=6000, data=data,files=files)
    #print(response.json())
    logging.info("--解析----"+str(response.json()))
    localFileJSON=wavFileName.lower().replace('.wav','.json')
    with open(localFileJSON, 'w') as jsonfile:
        jsonfile.write(str(response.json()).replace('\'','\"'))  # 将内容写入文件

    logging.info("--结束----"+mp4FileName) 

输出的文本json格式如下：

{"code": 0, "data": [{"end_time": "00:00:02,600", "line": 1, "start_time": "00:00:00,000", "text": "一分部署 九分落实"}, {"end_time": "00:00:06,400", "line": 2, "start_time": "00:00:02,600", "text": "聚焦16个县市区 开发区 保税区"}, {"end_time": "00:00:08,600", "line": 3, "start_time": "00:00:06,400", "text": "在装落实 求突破"}, {"end_time": "00:00:12,800", "line": 4, "start_time": "00:00:08,600", "text": "拼经济上的新思路 新作为 新使命"}, {"end_time": "00:00:16,400", "line": 5, "start_time": "00:00:12,800", "text": "展现全市各发展主体 努力建设"}, {"end_time": "00:00:18,200", "line": 6, "start_time": "00:00:16,400", "text": "实力强 品质优"}, {"end_time": "00:00:21,400", "line": 7, "start_time": "00:00:18,200", "text": "生活美的 更好为方的澎湃力量"}, {"end_time": "00:00:25,400", "line": 8, "start_time": "00:00:21,400", "text": "为方市广播电视台 大型柔媒体访谈节目"}, {"end_time": "00:00:29,600", "line": 9, "start_time": "00:00:25,400", "text": "落实进行时 已先后制作播出了六期节目"}, {"end_time": "00:00:33,600", "line": 10, "start_time": "00:00:29,600", "text": "在全市广大干部群众中引起强烈反响"}, {"end_time": "00:00:37,360", "line": 11, "start_time": "00:00:33,600", "text": "全市上下掀起了围绕七个加利突破"}, {"end_time": "00:00:40,960", "line": 12, "start_time": "00:00:37,360", "text": "比学赶超 争先进位的更大热潮"}, {"end_time": "00:00:43,960", "line": 13, "start_time": "00:00:40,960", "text": "落实进行时 重点抓落实"}, {"end_time": "00:00:47,560", "line": 14, "start_time": "00:00:43,960", "text": "第七期将播出特别节目 落实回头看"}, {"end_time": "00:00:51,400", "line": 15, "start_time": "00:00:47,560", "text": "对相关现实区主要工作推进落实情况"}, {"end_time": "00:00:53,160", "line": 16, "start_time": "00:00:51,400", "text": "进行调查采访"}, {"end_time": "00:00:55,200", "line": 17, "start_time": "00:00:53,160", "text": "进一步激发全市上下"}, {"end_time": "00:00:58,960", "line": 18, "start_time": "00:00:55,200", "text": "真抓实干 求突破的更大热情"}], "msg": "ok"}

输出的json文本中，带文本对应视频的开始时间和结束时间，时间精度为毫秒。

完整代码下载地址 Python实现从MP4文件中的语音提取文本 

有不清楚的地方，欢迎留言讨论。