ffmpeg使用总结

时间 2019-11-22

标签 ffmpeg 使用总结繁體版

原文原文链接

下载安装

提供了可执行文件和动态库，纯命令行使用Architecture版本。python

查看当前dshow可用的录像设备和录音设备

使用以下命令能够查看设备列表

ffmpeg -list_devices true -f dshow -i dummy

能够从输出中提取到设备名，进行视频录制时会使用到这个设备名shell

[dshow @ 0000023a33b3a680] DirectShow video devices (some may be both video and audio devices)
[dshow @ 0000023a33b3a680]  "USB 2.0 Webcam Device"
[dshow @ 0000023a33b3a680]     Alternative name "@device_pnp_\\?xxxx}\global"
[dshow @ 0000023a33b3a680]  "screen-capture-recorder"
[dshow @ 0000023a33b3a680]     Alternative name "@device_sw_{xxx}"
[dshow @ 0000023a33b3a680] DirectShow audio devices
[dshow @ 0000023a33b3a680]  "麦克风 (Realtek High Definition Audio)"
[dshow @ 0000023a33b3a680]     Alternative name "@device_cm_{xxx}"
[dshow @ 0000023a33b3a680]  "virtual-audio-capturer"
[dshow @ 0000023a33b3a680]     Alternative name "@device_sw_{xxxx}"

其中的 "screen-capture-recorder"和"virtual-audio-capturer"两个是虚拟设备，用来辅助录制屏幕和声卡， (下载地址)[https://sourceforge.net/projects/screencapturer/files/]。编程

查看单个设备的可用配置能够用以下命令

ffmpeg -list_options true -f dshow -i video="{录像设备名}"

能够从输出中获得摄像头的可用分辨率app

[dshow @ 000001d59c14a680] DirectShow video device options (from video devices)
[dshow @ 000001d59c14a680]  Pin "捕获" (alternative pin name "0")
[dshow @ 000001d59c14a680]   pixel_format=yuyv422  min s=640x480 fps=5 max s=640x480 fps=30
[dshow @ 000001d59c14a680]   pixel_format=yuyv422  min s=640x480 fps=5 max s=640x480 fps=30
...
[dshow @ 000001d59c14a680]   pixel_format=yuyv422  min s=1280x720 fps=5 max s=1280x720 fps=10
[dshow @ 000001d59c14a680]   vcodec=mjpeg  min s=640x480 fps=5 max s=640x480 fps=30
[dshow @ 000001d59c14a680]   vcodec=mjpeg  min s=640x480 fps=5 max s=640x480 fps=30
...
[dshow @ 000001d59c14a680]   vcodec=mjpeg  min s=1280x720 fps=5 max s=1280x720 fps=30
2018-03-19 13:39:48,178 - video=USB 2.0 Webcam Device: Immediate exit requested

获取桌面信息

ffmpeg -list_options true -f gdigrab -i desktop

能够从输出中提取出桌面分辨率ide

[gdigrab @ 000001f1b84da680] Capturing whole desktop as 1920x1080x32 at (0,0)

python代码提取设备列表和配置信息

本身封装的一个函数来获取设备可用分辨率列表函数式编程

__author__ = "Dalton Xiong"
__license__ = "GPL"
__version__ = "0.1"
__email__ = "daltonxiong@gmail.com"
'''
使用ffmpeg读取录像设备列表及其支持的分辨率
'''


import subprocess
import re
import logging
from collections import namedtuple

import parse

logging.basicConfig(level=logging.INFO,format='%(asctime)s - %(message)s')

# 读取设备列表， 忽略ffmpeg的录屏和录声卡的插件程序
def device_list():
    cmd = 'ffmpeg -list_devices true -f dshow -i dummy'
    logging.info('读取设备列表>>>>>>')
    logging.info(cmd)

    process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    output = process.stdout.read()
    lines = str(output, encoding='utf-8').split('\n')

    video_device_list = []
    audio_device_list = []
    video_device_name_flag = False
    audio_device_name_flag = False
    for l in lines:
        logging.info(l)
        l = l.strip()
        if '[dshow' in l:
            if 'DirectShow video devices' in l:
                video_device_name_flag = True
                audio_device_name_flag = False
            elif 'DirectShow audio devices' in l:
                video_device_name_flag = False
                audio_device_name_flag = True
            elif video_device_name_flag and 'Alternative name' not in l:
                video_device_name = re.search(r'\"(.*)\"', l) or None
                video_device_name = video_device_name and video_device_name.groups()[0]
                if video_device_name != 'screen-capture-recorder': # 忽略录屏程序
                    video_device_list.append(video_device_name)
            elif audio_device_name_flag and 'Alternative name' not in l:
                audio_device_name = re.search(r'\"(.*)\"', l) or None
                audio_device_name = audio_device_name and audio_device_name.groups()[0]
                if audio_device_name != 'virtual-audio-capturer': # 忽略声卡录制程序
                    audio_device_list.append(audio_device_name)

    return video_device_list, audio_device_list

# 影像采集设备信息
def video_device_info(device_name):
    cmd = 'ffmpeg -list_options true -f dshow -i video="{}"'.format(device_name)
    logging.info('读取影像采集设备({})的信息>>>>>>'.format(device_name))
    logging.info(cmd)

    process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    output = process.stdout.read()
    lines = str(output, encoding='utf-8').split('\n')

    pixel_format_set = set()
    vcodec_set = set()
    for l in lines:
        logging.info(l)
        l = l.strip()
        if '[dshow' not in l:
            continue

        if 'pixel_format' in l:
            result = parse.parse('{}   pixel_format={}  min s={} fps={} max s={}x{} fps={}', l)
            if result:
                pixel_format_set.add( (result[1], result[4], result[5], result[6]) )
        elif 'vcodec' in l:
            result = parse.parse('{}   vcodec={}  min s={} fps={} max s={}x{} fps={}', l)
            if result:
                vcodec_set.add( (result[1], result[4], result[5], result[6]) )
        else:
            continue

    return pixel_format_set, vcodec_set

# 声音采集设备信息
def audio_device_info(device_name):
    cmd = 'ffmpeg -list_options true -f dshow -i audio="{}"'.format(device_name)
    logging.info('读取声音采集设备({})的信息>>>>>>'.format(device_name))
    logging.info(cmd)

    process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    output = process.stdout.read()
    lines = str(output, encoding='utf-8').split('\n')

    format_set = set()
    for l in lines:
        logging.info(l)
        l = l.strip()
        if '[dshow' not in l:
            continue

        result = parse.parse('{}   min ch={} bits={} rate= {} max ch={} bits={} rate= {}', l)
        if result:
            format_set.add((result[4], result[5], result[6]))

    return format_set

# 桌面分辨率大小
def desktop_device_info():
    cmd = 'ffmpeg -list_options true -f gdigrab -i desktop'
    logging.info('读取桌面影像信息>>>>>>')
    logging.info(cmd)

    process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    output = process.stdout.read()
    lines = str(output, encoding='utf-8').split('\n')

    format_set = set()
    for l in lines:
        logging.info(l)
        l = l.strip()

        if '[gdigrab' not in l:
            continue

        result = parse.parse('{} Capturing whole desktop as {}x{}x{} at (0,0)', l)
        if result:
            format_set.add( (result[1], result[2]) )

    return format_set

# 获取全部设备列表和可用配置信息
def device_config_list():
    video_device_list, audio_device_list = device_list()

    # 录像设备
    VideoDeviceInfo = namedtuple('VideoDeviceInfo', ['width', 'height', 'fps'])
    vdevice_list = []
    for device_name in video_device_list:
        pixel_format_set, vcodec_set = video_device_info(device_name)
        format_set = vcodec_set or pixel_format_set
        format_list = set()
        for (format, width, height, fps) in format_set:
            format_list.add( VideoDeviceInfo(width=int(width), height=int(height), fps=int(fps)) )

        format_list = list(format_list)
        format_list.sort(key=lambda x: (x.fps, x.width), reverse=True)

        device_info = {'name': device_name, 'format_list': format_list}
        vdevice_list.append(device_info)

    # 录音设备
    AudioDeviceInfo = namedtuple('AudioDeviceInfo', ['channel', 'bits', 'rate'])
    adevice_list = []
    for device_name in audio_device_list:
        format_set = audio_device_info(device_name)
        format_list = set()
        for (channel, bits, rate) in format_set:
            format_list.add( AudioDeviceInfo(channel=int(channel), bits=int(bits), rate=int(rate)) )

        format_list = list(format_list)
        format_list.sort(key=lambda x: (x.channel, x.bits), reverse=True)

        device_info = {'name': device_name, 'format_list': format_list}
        adevice_list.append(device_info)

    # 计算机桌面
    desktop_format_set =  desktop_device_info()
    if len(desktop_format_set) != 1:
        raise ValueError('屏幕分辨率读取失败')

    desktop_format_set = list(desktop_format_set)[0]
    DesktopDeviceInfo = namedtuple('DesktopDeviceInfo', ['width', 'height'])
    desktop_info = DesktopDeviceInfo(width=int(desktop_format_set[0]), height=int(desktop_format_set[1]))

    logging.info('录像设备列表和可用配置以下：')
    logging.info('{}'.format(vdevice_list))

    logging.info('录音设备列表和可用配置以下：')
    logging.info('{}'.format(adevice_list))

    logging.info('桌面分辨率为：{}x{}'.format(desktop_info.width, desktop_info.height))

    return vdevice_list,adevice_list,desktop_info

if __name__ == '__main__':
    vdevice_list, adevice_list, desktop_info = device_config_list()

封装了一个函数device_config_list返回了，录像设备列表，录音设备列表，桌面分辨率函数

录像设备列表和可用配置以下：
[{'name': 'USB 2.0 Webcam Device', 'format_list': [VideoDeviceInfo(width=1280, height=720, fps=30), VideoDeviceInfo(width=960, height=540, fps=30), VideoDeviceInfo(width=848, height=480, fps=30), VideoDeviceInfo(width=640, height=360, fps=30), VideoDeviceInfo(width=640, height=480, fps=30), VideoDeviceInfo(width=424, height=240, fps=30), VideoDeviceInfo(width=352, height=288, fps=30), VideoDeviceInfo(width=320, height=240, fps=30), VideoDeviceInfo(width=320, height=180, fps=30), VideoDeviceInfo(width=176, height=144, fps=30), VideoDeviceInfo(width=160, height=120, fps=30)]}]
录音设备列表和可用配置以下：
[{'name': '麦克风 (Realtek High Definition Audio)', 'format_list': [AudioDeviceInfo(channel=2, bits=16, rate=44100)]}]
桌面分辨率为：1920x1080

其余

~~ffmpeg -list_devices true -f gdigrab -i dummy~~动画

把dshow换成gdigrab。。。能够看到提示让你使用desktop或者对应的窗口名ui

采集摄像头和桌面和麦克风

使用python启动和中止录制进程

启动ffmpeg子进程

ffmpeg_process = subprocess.Popen('ffmpeg -f gdigrab -i desktop -y out.mp4', shell=True, stdin=subprocess.PIPE)

中止子进程,ffmpeg命令行收到'q'就会自动退出

ffmpeg_process.communicate(b'q')

效果1

在不变更桌面和摄像头的分辨率状况下，让桌面画面和摄像头并排显示，hstack可让两个画面并排显示，hstack函数要求两边的画面的高度必须保持一致，所以加上pad函数来对高度短的画面填充纯色。

代码分段加注释以后以下

fmpeg 
-f gdigrab -i desktop # 输入桌面画面
-f dshow -i video="USB 2.0 Webcam Device" -s 1280*720 # 输入摄像头画面
-filter_complex "[0:v]pad=1920:1080:0:0:black[left];[1:v]pad=1280:1080:0:0:black[right];[left][right]hstack"  #滤镜操做
-f dshow -i audio="麦克风 (Realtek High Definition Audio)"  # 输入麦克风
-y out.mp4  # 输出到文件

语法解析

-f gdigrab -i desktop 获取桌面画面， -f设置输入格式， -i设置输入源
-f dshow -i video="{设备名}" -s {分辨率设置} 获取摄像头画面，设备名是从上面提取到的，分辨率也是查看设备配置获得的，这里使用的是个人摄像头的支持的最高分辨率，格式是{宽}*{高}
-filter_complex, 滤镜处理，处理流程用双引号包起来了，里面的语法是相似函数式编程, 单纯的输入数据处理而后输出数据，[0:v]和[1:v] 分别表示前面两个用-i引入的第一个和第二个画面源。第一个分号语句[0:v]pad=1920:1080:0:0:black[left];第一个中括号表示输入，末尾中括号表示输出，中间是调用pad函数，参数列表用冒号隔开，每一个位置的参数意义能够查看文档。第二个分号语句[1:v]pad=1280:1080:0:0:black[right];同上，最后一段是[left][right]hstack末尾没有中括号和分号，表示输出滤镜的最后结果。
-f dshow -i audio="麦克风 (Realtek High Definition Audio)" 获取麦克风数据
-y out.mp4 将滤镜的画面结果和麦克风的声音数据合并，输出到文件out.mp4, -y表示若是文件存在就替换。

滤镜函数文档

效果2

仍是使用hstack来并排显示两个画面，仍然须要将两个画面的高度调成一致，这边使用scale函数来修改画面的分辨率，将两个画面的高度调为二者中的高值。

python组装命令行代码

cmd = '''
    ffmpeg
    -f gdigrab -i desktop
    -f dshow -i video="{vname}" -s {vwidth}*{vheight}
    -filter_complex "[0:v]scale=-1:{max_height}[left];[1:v]scale=-1:{max_height}[right];[left][right]hstack,scale=1980:-1"
    -f dshow -i audio="{aname}"
    -y out.mp4
'''.format(
    vname=vdevice['name'],
    vwidth=vdevice['format_list'][0].width,
    vheight=vdevice['format_list'][0].height,
    max_height = max(desktop_info.height, vdevice['format_list'][0].height),
    aname=adevice['name']
)

cmd = cmd.replace('\n', ' ').replace('   ', ' ').replace('   ', ' ').replace('   ', ' ')   #删除富余的空格

个人机器上最后执行的命令是

ffmpeg -f gdigrab -i desktop -f dshow -i video="USB 2.0 Webcam Device" -s 1280*720 -filter_complex "[0:v]scale=-1:1080[left];[1:v]scale=-1:1080[right];[left][right]hstack,scale=1980:-1" -f dshow -i audio="麦克风 (Realtek High Definition Audio)" -y out.mp4

语法解析

[0:v]scale=-1:{max_height}[left];，scale调整分辨率大小，将高度设置为二者的最高值，第一个参数为-1，表示按比例变化。
[left][right]hstack,scale=1980:-1, 最后合并的画面将长度改为了1980高度成比例变化，逗号语法将hstack的输出直接做为scale的输入，省去了中间变量的命名。

效果3

让摄像头画面浮动到桌面的右下角可使用overlay方法

python组装命令行代码

cmd = '''
    ffmpeg
    -f gdigrab -i desktop
    -f dshow -i video="{vname}" -s {vwidth}*{vheight}
    -filter_complex "[0:v][1:v]overlay=main_w-overlay_w:main_h-overlay_h"
    -f dshow -i audio="{aname}"
    -y out.mp4
'''.format(
    vname=vdevice['name'],
    vwidth=vdevice['format_list'][0].width,
    vheight=vdevice['format_list'][0].height,
    max_height = max(desktop_info.height, vdevice['format_list'][0].height),
    aname=adevice['name']
)

cmd = cmd.replace('\n', ' ').replace('   ', ' ').replace('   ', ' ').replace('   ', ' ')   #删除富余的空格

个人机器上最后执行的命令是

ffmpeg -f gdigrab -i desktop -f dshow -i video="USB 2.0 Webcam Device" -s 1280*720 -filter_complex "[0:v][1:v]overlay=main_w-overlay_w:main_h-overlay_h" -f dshow -i audio="麦克风 (Realtek High Definition Audio)"  -y o.mp4

语法解析

[0:v][1:v]overlay=main_w-overlay_w:main_h-overlay_h, overlay方法接收两个输入参数，第一个是[0:v]底层画面，这边使用的是桌面，第二个是[1:v]浮动画面，这边使用的是摄像头画面, overlay设置浮动画面相对于底层画面的偏移位置坐标，main_w,overlay_w,main_h,overlay_h是经过两个输入参数的附加参数，其余附加参数能够看文档, main_w和main_h是底层画面的宽度和高度，overlay_w和overlay_h是浮动画面的宽度和高度，坐标(main_w-overlay_w, main_h-overlay_h)正好就能够将浮动画面完美的放置到右下角。