Recherche avancée

Sur d’autres sites (9809)

• How can I stream raw video frames AND audio to FFMPEG with Python 2.7 ?

  18 novembre 2017, par Just Askin

  I am streaming raw video frames from Pygame to FFMPEG, then sending to a rtmp stream, but for the life of me, I can’t figure out how to send live audio using the same Python module. It does not need to be the Pygame mixer, but I am not opposed to using it if that is where the best answer lies. I’m pretty sure it’s not though.

  My question is this : What is the best strategy to send live audio output from a program to FFMPEG along with raw video frames simultaneously from the same Python module ?

  My program is large, and eventually I would like to build options to switch audio inputs from a queue of music, a microphone, or any other random sounds from any program I want to use. But for the time being, I just want something to work. I am starting off with a simple Espeak command.

  Here is my Python commands :

  command = ['ffmpeg', '-re', '-framerate', '22', '-s', '1280x720', '-pix_fmt', 'rgba', '-f', 'rawvideo', '-i', '-', '-f', 's16le', '-ar', '22500', '-i', '/tmp/audio', '-preset', ultrafast', '-pix_fmt', 'rgba', '-b:v', '2500', '-s', 'hd720', '-r', '25', '-g', '50', '-crf', '20', '-f', 'flv', 'rtmp://xxx' ]
  
  
  
  pipe = sp.Popen(command, stdin=sp.PIPE) 

  Then I send my frames to stdin from within my main while True: loop.

  The problem I run into with this strategy is I can’t figure out how to shove audio into FFMPEG from within Python without blocking the pipe. After hours of research, I am pretty confident I can’t use the pipe to send the audio along with the frames. I thought the named pipe was my solution (which works running Espeak outside of Python), but it blocks Python until the Espeak is done... so no good.

  I assume I need threading for multiprocessing, but I cannot figure out from the official documentation or any other resources as to how I can solve my problem with it.

  The ['-f', 's16le', '-ar', '22500', '-i', '/tmp/audio'] are settings that work if I run espeak from a separate terminal with espeak 'some text' --stdout > /tmp/audio.

  I am using Centos 7, Python 2.7, pygame, the latest build of FFMPEG,

• OpenCV Alpha Channel support

  14 février 2014, par adriagil

  I've tried many different solutions but I'm stuck at this point.

  I have a sequence of .png files with alpha channel.
  If I pick one of the files for splitting the channels I got the expected result in an array[4] having the alpha channel

  Mat check = imread("1.png");

  printf("channels = %d", check.channels()); //got 'channels = 4'

  Then I expected to get the same results for a movie file.

  With FFMPEG I've just converted the .png sequence to a .mov file with "qtrle" codec that I'm sure that support alpha channel.

  ffmpeg -pix_fmt argb -i sequence_%d.png -vcodec qtrle output.mov

  Then the I process the video file frames with OpenCV

  Mat frame;
  
  VideoCapture cap ("output.mov");
  
  if (cap.grab())
  
      cap.retrieve(frame);
  
  printf("channels = %d", frame.channels()); // got 'channels = 3'

  I've checked the ffmpeg generated output and seems to be encoded right and have the alpha channel stored.

  Does OpenCV does not support Alpha Channel in movie files ?
  If so, anyone knows an alternative to do it with C++ or other libraries ?
  Can this be done with DirectX in some way (only using OpenCV for reading video) ?

  In the official docs I've found that cv::VideocCapture.retrieve() has a second argument for the 'channel' but I've tried to do the following with the same results (no alpha channel) :

  cap.retrieve(frame, 4);
  
  cap.retrieve(frame, -1);

  As far as cv::VideoCapture supports loading image sequences I've tried to load the PNG sequence but I got the following warning so I could not play the movie file :

  VideoCapture cap("sequence_%d.png");
  
  warning: Could not find codec parameters (../../modules/highgui/src/cap_ffmpeg_impl.hpp)

  Why I got that result if I can read the same PNG with imread("") ?

  Also I've tried to encode the .png sequence again with ffmpeg :

  ffmpeg -pix_fmt rgba -i sequence_%d.png -vcodec png output.mov

  But got exactly the same warning as before.

  Any suggestion would be much appreciated !

  Note : I'm using OpenCV 2.4.2 right now...maybe updating to 2.4.8 may solve the problem ?

• The question about ffmpeg drawtext filter [closed]

  5 mai 2024, par B1GGersnow

  I tried to use rockchip(aarch64) hardware acceleration and add a drawtext filter to add watermarks. However, squares appear when multiple Chinese fonts are added.

  


  This is my compilation parameter.

  


  ./configure --prefix=/usr --enable-gpl --enable-version3 --enable-libdrm --enable-rkmpp --enable-rkrga --enable-filter=drawtext --enable-libharfbuzz --enable-libfreetype --enable-libfontconfig --enable-libfreetype --enable-libfribidi


  


  


  1. ffmpeg -hwaccel rkmpp -hwaccel_output_format drm_prime -i 1.mp4 -vf scale_rkrga=w=1920:h=1080,hwdownload,format=nv12,drawtext=text='中文':fontfile=msyh.ttc:fontsize=200 -c:v h264_rkmpp -y -t 10  output.mp4
img1

     


  


  2. ffmpeg -hwaccel rkmpp -hwaccel_output_format drm_prime -i 1.mp4 -vf "scale_rkrga=w=1920:h=1080,hwdownload,format=nv12,drawtext=text='中文字幕测试':fontfile=msyh.ttc:fontsize=200" -c:v h264_rkmpp -y -t 10  output.mp4
img2

     


  


  


  But I try to use apt install ffmpeg to install ffmpeg which is officially maintained by Ubuntu. I got the right result. So I think it's the library, and I'm trying to compile using the official library.

  


  ./configure --prefix=/usr --enable-gpl --enable-version3 --enable-filter=drawtext --enable-libharfbuzz --enable-libfreetype --enable-libfontconfig --enable-libfreetype --enable-libfribidi --enable-libx264


  


  I still get garbled results.

  


  Is this because there is something wrong with my compilation dependent library ?

  


  When I compiled it with the same parameters on wsl, there was no garbled code.