Recherche avancée

Sur d’autres sites (7677)

• Inputting Audio Stream to FFMPEG

  5 avril 2018, par Wijaya

  I’m building a real time chat application with C# and ffmpeg.exe. My requirement is to get a memory stream from Microsoft Speech API and feed it to ffmpeg process in real time. I can take a memory stream from Microsoft Speech API. I’m using following code to create the memory stream.

          using (MemoryStream stream = new MemoryStream())
  
          {
  
              MemoryStream streamAudio = new MemoryStream();
  
              System.Media.SoundPlayer m_SoundPlayer = new System.Media.SoundPlayer();
  
              _speechSynthesizerVisemesSender.SetOutputToWaveStream(streamAudio);
  
              _speechSynthesizerVisemesSender.SetOutputToNull();
  
              stream.WriteTo(proc.StandardInput.BaseStream);
  
          }   

  I’m already using another datapipe with another command to feed video content to ffmpeg. But I couldn’t find a stable solution to feed audio through a datapipe. This article briefly explains about audio datapipe. I’m using following command to stream audio.

  "ffmpeg -re -f s16le -i pipe:wav -f mpegts udp://127.0.0.1:1234"

  But it is not working with the datapipe. If I try the command with mp3 or wav file, it works.

• python - subprocess.Popen fails to correctly pipe urllib3 Response

  11 février 2018, par user2963567

  I want to open a file over a network and pipe it directly to ffmpeg using subprocess.Popen. The goal is to stream the audio file data directly into ffmpeg. Here is the code :

  # test.py
  
  import subprocess, sys, urllib3, time
  
  
  
  http = urllib3.PoolManager()
  
  r = http.urlopen('GET', sys.argv[1], preload_content=False)
  
  
  
  args = 'ffmpeg -i - -y audio.mp3'.split(' ')
  
  subprocess.Popen(args, stdin=r)
  
  
  
  r.close()

  If I run a local HTTP server and give the program the url, it works successfully, and ffmpeg processes it.

  $ python3 test.py http://192.168.1.200/original.webm

  However if I try to retrieve from a remote server, such as that below, ffmpeg fails.

  $ python3 test.py https://cdn.discordapp.com/attachments/304959901376053248/412003156638040084/original.webm

  with the following output

  pipe:: Invalid data found when processing input

  I expected this code to produce the same results as running this terminal command. This command succeeds for both the discord cdn URL and a local HTTP server url.

  $ curl [file url] | ffmpeg -i - -y audio.mp3

  I’m using python 3.5 on Linux, and ffmpeg 3.4.1.

  edit 1

  I’m now leaning towards thinking it’s not ffmpeg’s fault, and more about how Popen is reading/writing the urllib response to a process’ stdin. By running a local netcat server and sending the output to a file ($ nc -l 1234 > nc_output.webm) and adjusting the script like so :

  import subprocess, sys, urllib3, time
  
  
  
  http = urllib3.PoolManager()
  
  r = http.urlopen('GET', sys.argv[1], preload_content=False)
  
  
  
  args = 'nc 192.168.1.200 1234'.split(' ')
  
  subprocess.run(args, stdin=r)
  
  
  
  r.close()

  Then running as $ python3 test.py https://cdn.discordapp.com/attachments/304959901376053248/412003156638040084/original.webm

  By comparing nc_output.webm with the original.webm file, I can immediately see that nc_output.webm is slightly larger (4017585 bytes, vs 4008589 bytes). Attempting to play nc_output.webm (mpv, vlc, ffprobe) also fails, which explains why ffmpeg was complaining. Whatever Popen is doing to the stream’s bytes is sufficient to make the output file useless.

  However, the problem still ceases to occur if the URL points to a local HTTP server, such as one run from python -m SimpleHTTPServer which leads me to think that this is related to the latency associated with reading from a remote origin.

• Does not work -hls_playlist in ffmpeg in the -dash section

  18 mars 2019, par jidckii

  I will use this assembly ffmpeg
  https://hub.docker.com/r/jrottenberg/ffmpeg/

  ffmpeg version 3.4.1 Copyright (c) 2000-2017 the FFmpeg developers
  
  built with gcc 6.2.1 (Alpine 6.2.1) 20160822  

  I’m trying to do dash and hls at the same time.

  In the documentation from here http://ffmpeg.org/ffmpeg-all.html#dash-2

  the key is :

  -hls_playlist hls_playlist
  
  
  
  Generate HLS playlist files as well. The master playlist is generated
  
  with the filename master.m3u8. One media playlist file is generated for
  
  each stream with filenames media_0.m3u8, media_1.m3u8, etc.

  those. mpd and m3u8 must be created at the same time.

  Running :

  ffmpeg -re \
  
  -analyzeduration 20000000 \
  
  -i udp: //239.0.0.1: 1234? overrun_nonfatal = 1 \
  
  -map 0 \
  
  -c copy \
  
  -f dash \
  
  -min_seg_duration 5000000 \
  
  -window_size 10 \
  
  -extra_window_size 10 \
  
  -remove_at_exit 1 \
  
  -hls_playlist 1 \
  
  -use_timeline 1 \
  
  -use_template 1 \
  
  -vtag avc1 \
  
  -atag mp4a \
  
  cam01.mpd

  and get an error :

  Unrecognized option 'hls_playlist'.  
  
  Error splitting the argument list: Option not found`

  without the -hls_playlist
  everything works correctly ...

  What could be the problem ?