下載地址 html
提供了可執行文件和動態庫,純命令行使用Architecture版本。python
ffmpeg -list_devices true -f dshow -i dummy
能夠從輸出中提取到設備名,進行視頻錄製時會使用到這個設備名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] "麥克風 (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"兩個是虛擬設備,用來輔助錄製屏幕和聲卡, (下載地址)[https://sourceforge.net/projects/screencapturer/files/]。編程
ffmpeg -list_options true -f dshow -i video="{錄像設備名}"
能夠從輸出中獲得攝像頭的可用分辨率app
[dshow @ 000001d59c14a680] DirectShow video device options (from video devices) [dshow @ 000001d59c14a680] Pin "捕獲" (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
ffmpeg -list_options true -f gdigrab -i desktop
能夠從輸出中提取出桌面分辨率ide
[gdigrab @ 000001f1b84da680] Capturing whole desktop as 1920x1080x32 at (0,0)
本身封裝的一個函數來獲取設備可用分辨率列表函數式編程
__author__ = "Dalton Xiong" __license__ = "GPL" __version__ = "0.1" __email__ = "daltonxiong@gmail.com" ''' 使用ffmpeg讀取錄像設備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') # 讀取設備列表, 忽略ffmpeg的錄屏和錄聲卡的插件程序 def device_list(): cmd = 'ffmpeg -list_devices true -f dshow -i dummy' logging.info('讀取設備列表>>>>>>') 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': # 忽略錄屏程序 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': # 忽略聲卡錄製程序 audio_device_list.append(audio_device_name) return video_device_list, audio_device_list # 影像採集設備信息 def video_device_info(device_name): cmd = 'ffmpeg -list_options true -f dshow -i video="{}"'.format(device_name) logging.info('讀取影像採集設備({})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 # 聲音採集設備信息 def audio_device_info(device_name): cmd = 'ffmpeg -list_options true -f dshow -i audio="{}"'.format(device_name) logging.info('讀取聲音採集設備({})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('讀取桌面影像信息>>>>>>') 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 # 獲取全部設備列表和可用配置信息 def device_config_list(): video_device_list, audio_device_list = device_list() # 錄像設備 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) # 錄音設備 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) # 計算機桌面 desktop_format_set = desktop_device_info() if len(desktop_format_set) != 1: raise ValueError('屏幕分辨率讀取失敗') 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('錄像設備列表和可用配置以下:') logging.info('{}'.format(vdevice_list)) logging.info('錄音設備列表和可用配置以下:') logging.info('{}'.format(adevice_list)) logging.info('桌面分辨率爲:{}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()
封裝了一個函數device_config_list
返回了, 錄像設備列表,錄音設備列表, 桌面分辨率函數
錄像設備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)]}] 錄音設備列表和可用配置以下: [{'name': '麥克風 (Realtek High Definition Audio)', 'format_list': [AudioDeviceInfo(channel=2, bits=16, rate=44100)]}] 桌面分辨率爲:1920x1080
ffmpeg -list_devices true -f gdigrab -i dummy動畫
把dshow換成gdigrab。。。能夠看到提示讓你使用desktop或者對應的窗口名ui
啓動ffmpeg子進程
ffmpeg_process = subprocess.Popen('ffmpeg -f gdigrab -i desktop -y out.mp4', shell=True, stdin=subprocess.PIPE)
中止子進程,ffmpeg命令行收到'q'就會自動退出
ffmpeg_process.communicate(b'q')
在不變更桌面和攝像頭的分辨率狀況下,讓桌面畫面和攝像頭並排顯示,hstack可讓兩個畫面並排顯示,hstack函數要求兩邊的畫面的高度必須保持一致,所以加上pad函數來對高度短的畫面填充純色。
代碼分段加註釋以後以下
fmpeg -f gdigrab -i desktop # 輸入桌面畫面 -f dshow -i video="USB 2.0 Webcam Device" -s 1280*720 # 輸入攝像頭畫面 -filter_complex "[0:v]pad=1920:1080:0:0:black[left];[1:v]pad=1280:1080:0:0:black[right];[left][right]hstack" #濾鏡操做 -f dshow -i audio="麥克風 (Realtek High Definition Audio)" # 輸入麥克風 -y out.mp4 # 輸出到文件
-f gdigrab -i desktop
獲取桌面畫面, -f
設置輸入格式, -i
設置輸入源-f dshow -i video="{設備名}" -s {分辨率設置}
獲取攝像頭畫面, 設備名是從上面提取到的,分辨率也是查看設備配置獲得的,這裏使用的是個人攝像頭的支持的最高分辨率,格式是{寬}*{高}
-filter_complex
, 濾鏡處理,處理流程用雙引號包起來了,裏面的語法是相似函數式編程, 單純的輸入數據處理而後輸出數據,[0:v]
和[1:v]
分別表示前面兩個用-i
引入的第一個和第二個畫面源。第一個分號語句[0:v]pad=1920:1080:0:0:black[left];
第一個中括號表示輸入,末尾中括號表示輸出,中間是調用pad函數,參數列表用冒號隔開,每一個位置的參數意義能夠查看文檔。第二個分號語句[1:v]pad=1280:1080:0:0:black[right];
同上,最後一段是[left][right]hstack
末尾沒有中括號和分號,表示輸出濾鏡的最後結果。-f dshow -i audio="麥克風 (Realtek High Definition Audio)"
獲取麥克風數據-y out.mp4
將濾鏡的畫面結果和麥克風的聲音數據合併,輸出到文件out.mp4, -y
表示若是文件存在就替換。仍是使用hstack來並排顯示兩個畫面,仍然須要將兩個畫面的高度調成一致,這邊使用scale函數來修改畫面的分辨率,將兩個畫面的高度調爲二者中的高值。
python組裝命令行代碼
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(' ', ' ') #刪除富餘的空格
個人機器上最後執行的命令是
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="麥克風 (Realtek High Definition Audio)" -y out.mp4
[0:v]scale=-1:{max_height}[left];
,scale調整分辨率大小,將高度設置爲二者的最高值,第一個參數爲-1,表示按比例變化。[left][right]hstack,scale=1980:-1
, 最後合併的畫面將長度改爲了1980高度成比例變化,逗號語法將hstack的輸出直接做爲scale的輸入,省去了中間變量的命名。讓攝像頭畫面浮動到桌面的右下角可使用overlay方法
python組裝命令行代碼
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(' ', ' ') #刪除富餘的空格
個人機器上最後執行的命令是
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="麥克風 (Realtek High Definition Audio)" -y o.mp4
[0:v][1:v]overlay=main_w-overlay_w:main_h-overlay_h
, overlay
方法接收兩個輸入參數,第一個是[0:v]
底層畫面,這邊使用的是桌面, 第二個是[1:v]
浮動畫面, 這邊使用的是攝像頭畫面, overlay
設置浮動畫面相對於底層畫面的偏移位置座標,main_w,overlay_w,main_h,overlay_h是經過兩個輸入參數的附加參數,其餘附加參數能夠看文檔, main_w和main_h是底層畫面的寬度和高度,overlay_w和overlay_h是浮動畫面的寬度和高度, 座標(main_w-overlay_w, main_h-overlay_h)正好就能夠將浮動畫面完美的放置到右下角。