Python 3调用FFmpeg子进程获取stderr时出现字符解码错误

问题根源 #

你遇到的'charmap' codec can't decode byte错误，本质是编码不匹配：

• 设置universal_newlines=True后，subprocess会使用系统默认编码（Windows下通常是GBK/GB2312）解码FFmpeg的stderr输出
• 但FFmpeg的标准错误输出默认采用UTF-8编码，日文字符的字节流用系统默认编码解码就会出现无法识别的字符，触发解码失败

解决方案 #

方案1：直接指定编码（Python 3.7+） #

Python 3.7及以上版本的subprocess.Popen支持encoding参数，直接指定为utf-8即可：

command = [ffmpeg, "-y", "-i", fileSelected, "-acodec", "pcm_s16le", "-vn", "-t", "3", "-f", "null", "-"]
print(command)   
# 替换universal_newlines=True为encoding='utf-8'，同时修正Jstdin的拼写错误
proc = subprocess.Popen(command, stderr=subprocess.PIPE, stdin=subprocess.PIPE, encoding='utf-8', startupinfo=startupinfo)
  
stream = ""    
for line in proc.stderr:
    try:
        print("line", line)
        if "Stream #" in line:
            estream = line.split("#",1)[1]
            estream = estream.split(" (",1)[0]
            print("estream", estream)
            stream += estream + "\n"
    except Exception as error:  # 修正exception为大写Exception
        print("处理错误:", error)

方案2：手动解码字节流（兼容Python 3.6及以下） #

如果你的Python版本低于3.7，不支持encoding参数，可以直接读取字节流，再手动用UTF-8解码：

command = [ffmpeg, "-y", "-i", fileSelected, "-acodec", "pcm_s16le", "-vn", "-t", "3", "-f", "null", "-"]
print(command)   
# 去掉universal_newlines=True，直接读取字节
proc = subprocess.Popen(command, stderr=subprocess.PIPE, stdin=subprocess.PIPE, startupinfo=startupinfo)
  
stream = ""    
for line_bytes in proc.stderr:
    try:
        # 用UTF-8解码字节流，strip()去除换行符
        line = line_bytes.decode('utf-8').strip()
        print("line", line)
        if "Stream #" in line:
            estream = line.split("#",1)[1]
            estream = estream.split(" (",1)[0]
            print("estream", estream)
            stream += estream + "\n"
    except UnicodeDecodeError as error:
        print("解码错误:", error)
        # 可选：用replace模式忽略无法解码的字符，避免程序中断
        line = line_bytes.decode('utf-8', errors='replace').strip()
        # 后续可继续处理line内容

额外注意事项 #

1. 修正原代码中的拼写错误：Jstdin应为stdin，exception应为大写Exception
2. 如果遇到极少数FFmpeg输出编码不是UTF-8的情况，可以尝试shift-jis编码（日文常见编码）进行解码

内容的提问来源于stack exchange，提问作者jdauthre