关于视频的处理，利用ffmpeg提取视频字幕和音频，如果ffmpeg无法识别字幕，则通过ocr识别字幕

ffmpeg提取视频字幕和音频，代码

import ffmpeg

input_video = r"D:\GZ\soft\PaddleOCR-release-2.7\MP4\11.mp4"
output_audio = "output_audio.aac"
output_subtitle = "output_subtitle.srt"

try:
    probe = ffmpeg.probe(input_video, v='error')

    # 检查是否有音频流
    audio_stream = next((stream for stream in probe['streams'] if stream['codec_type'] == 'audio'), None)
    if audio_stream:
        print("提取音频流...")
        ffmpeg.input(input_video).output(output_audio, codec='copy').run()
    else:
        print("没有音频流可提取")

    # 检查是否有字幕流
    subtitle_stream = next((stream for stream in probe['streams'] if stream['codec_type'] == 'subtitle'), None)
    if subtitle_stream:
        print("提取字幕流...")
        ffmpeg.input(input_video).output(output_subtitle, codec='srt').run()
    else:
        print("没有字幕流可提取")

except ffmpeg.Error as e:
    print(f"ffmpeg error:\n{e.stderr.decode('utf-8')}")

如果没有字幕，则用ocr提取字幕，我在github上找到两个相关案例

专注于字幕的

YaoFANGUK/video-subtitle-extractor: 视频硬字幕提取，生成srt文件。无需申请第三方API，本地实现文本识别。基于深度学习的视频字幕提取框架，包含字幕区域检测、字幕内容提取。A GUI tool for extracting hard-coded subtitle (hardsub) from videos and generating srt files. (github.com) 可以提取各种图片，文档，功能比教强大，但需要自己按需求修改的

PaddlePaddle/PaddleOCR: Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices) (github.com)