Recherche avancée

Sur d’autres sites (5172)

• Is there a way to create a single HLS output from more than one RTMP inputs in listen mode ?

  25 novembre 2020, par Gabriele Tassoni

  I'm trying to create a single HLS output using, as inputs, different RTMP playpaths or different RTMP streams, i.e. :

  


  

  • rtmp ://ip:1935/app/key0

  


  • rtmp ://ip:1935/app/key1

  


  • rtmp ://ip:1935/app/key2

  


  


  The resulting HLS output should have video track from key0, and three audio tracks from key0, key1 and key2.

  


  I used FFMPEG, but I'm open to different open source software for the server/converter part.

  


  What I tried so far :

  


  NO RTMP approach

  


  I get in input anything thrown to a TCP port, the mp4 stream sent already has multiple audio tracks, this works well and the HLS output is correctly created, cons : it's not compatible with RTMP standard, many commercial software, like vMix, just live streams to RTMP, so I cannot use this approach.

  


  ffmpeg -y -hwaccel auto \
    -i tcp://0.0.0.0:${port}?listen \
    -map 0:0 \
    -map 0:a \
    -c:v libx264 -c:a aac -b:v 5000k -b:a 192k -vf "scale=1920:trunc(ow/a/2)*2" -tune zerolatency -preset veryfast -crf 23 -g 48 \
      -f hls -hls_time 4 -hls_playlist_type event -hls_segment_filename ${dir}/${name}/1080p_%03d.ts ${dir}/${name}/1080p.m3u8


  


  NGINX RTMP Module

  


  I used in NGINX RTMP an application which execs an ffmpeg command, cons : it doesn't work, it sits there forever, like it's waiting for something.

  


  ffmpeg -y -hwaccel auto \
        -i "rtmp://localhost:1935/mainav/$key" \
        -i "rtmp://localhost:1935/mainav/$name:lang:$lang" \
        -map 0:v:0 -map 0:a:0 -metadata:s:a:0 "language=$default" \
        -map "1:a:0" -metadata:s:a:1 "language=$lang" \
        -c:v libx264 -c:a aac -b:v 5000k -b:a 192k -vf "scale=1920:trunc(ow/a/2)*2" -tune zerolatency -preset veryfast -crf 23 -g 48 \
        -f hls -hls_time 4 -hls_playlist_type event -hls_segment_filename "${dir}/${name}/1080p_%03d.ts" "${dir}/${name}/1080p.m3u8"



  


  FFMPEG Listener

  


  Quite similar to the NGINX version, but without NGINX, going raw on bare ffmpeg server capabilities. Cons : it doesn't work, just throws an I/O Error as soon as the client connects.

  


  ffmpeg -y -hwaccel auto \
            -listen 1 -i "rtmp://0.0.0.0:1935/mainav/$key" \
            -listen 1 -i "rtmp://0.0.0.0:1935/mainav/$name:lang:$lang" \
            -map 0:v:0 -map 0:a:0 -metadata:s:a:0 "language=$default" \
            -map "1:a:0" -metadata:s:a:1 "language=$lang" \
            -c:v libx264 -c:a aac -b:v 5000k -b:a 192k -vf "scale=1920:trunc(ow/a/2)*2" -tune zerolatency -preset veryfast -crf 23 -g 48 \
            -f hls -hls_time 4 -hls_playlist_type event -hls_segment_filename "${dir}/${name}/1080p_%03d.ts" "${dir}/${name}/1080p.m3u8"


  


  Client used for tests

  


  I used an ffmpeg which reads a file from my disk and sends to different RTMP playpaths :

  


  ffmpeg -y -hwaccel auto -re \
    -i /home/me/Videos/xm.mkv \
    -map 0:v:0 \
    -map 0:a:0 \
    -f flv rtmp://localhost:1935/mainav/thebiglive:default:ita \
    -map 0:a:1 \
    -f flv rtmp://localhost:1935/mainav/thebiglive:lang:eng


  


  I feel a bit clueless now. The only working solution is not a standard one taken into account the common live encoders found in the market. Anyone has a pointer for which direction I can take ?

  


  Thanks

  


• How to get the video count of youtube_dl (1 of 256) in its output and hide ffmpeg console when converting the format ?

  28 août 2020, par Kristen

  I am creating a GUI program in Qt creator via Pyside2 and it's a program that lets you download Youtube music. My goal is to get the video count like x of 256 or x of 2 or 1 of 256 when downloading the playlist but I don't know how. I've come across two problems.

  


  First problem : I don't know how to get this information from youtube_dl output

  


  [youtube:playlist] Downloading playlist PLi0XaxnSihZSnD1TNm82Y43o92r6UNEfH - add --no-playlist to just download video -5EmnQp3V48
[youtube:playlist] PLi0XaxnSihZSnD1TNm82Y43o92r6UNEfH: Downloading webpage
[download] Downloading playlist: Dance Central Songs
[youtube:playlist] PLi0XaxnSihZSnD1TNm82Y43o92r6UNEfH: Downloading page #1
[youtube:playlist] PLi0XaxnSihZSnD1TNm82Y43o92r6UNEfH: Downloading page #2
[youtube:playlist] playlist Dance Central Songs: Downloading 251 videos
[download] Downloading video **1 of 251**
[youtube] bESGLojNYSo: Downloading webpage
[download] Destination: Lady Gaga - Poker Face (Official Music Video)-bESGLojNYSo.webm
[download] 100% of 3.50MiB in 00:01                   
Finished
[ffmpeg] Destination: Lady Gaga - Poker Face (Official Music Video)-bESGLojNYSo.mp3
Deleting original file Lady Gaga - Poker Face (Official Music Video)-bESGLojNYSo.webm (pass -k to keep)
[download] Downloading video **2 of 251**
[youtube] E2tMV96xULk: Downloading webpage
14:03:36: C:\Python38\python.exe exited with code 0


  


  Second problem : If my program is in .exe format then FFMPEG console appears and disappears immediately after it has completed the conversion but it's taking the user focus every time which gets annoying.

  


  What I've tried for first problem is that I tried to get it manually but it did not work then I checked youtube_dl github download dictionary keys and there is not a key that can acquire that video count 1 of 256

  


  print("Download keys:", download.keys())
Download keys: dict_keys(['status', 'downloaded_bytes', 'total_bytes', 'tmpfilename', 'filename', 'eta', 'speed', 'elapsed', '_eta_str', '_percent_str', '_speed_str', '_total_bytes_str'])


  


  What I've tried for second problem is actually none. I checked C# versions and with C# you can hide that console or disable it but I haven't found a way to do that in Python yet. Goal here is disable ffmpeg console taking either the focus or hide/disable it or third option is to use another converter.

  


  Youtube_dl code that is in Worker thread :

  


  def run(self):
    self.main_window.progressBar.setValue(0)

    self.ydl_opts = {
        'format': 'bestaudio/best',
        'noplaylist': self.main_window.playlist,
        'source_address': '0.0.0.0',
        'ignoreerrors': True,
        'progress_hooks': [self.my_hook],
        'postprocessors': [{
            'key': 'FFmpegExtractAudio',
            'preferredcodec': 'mp3'
        }]
    }

    with youtube_dl.YoutubeDL(self.ydl_opts) as ydl:
        ydl.download([f'{self.main_window.downloadURL.text()}'])

def my_hook(self, download):
    if download["status"] == "finished":
        print("Finished")
        self.main_window.progressText.setText("Ready")
    if download["status"] == "downloading":
        self.updateText.emit("Downloading...")
        percent = download["_percent_str"]
        percent = percent.replace("%", "")
        self.updateProgress.emit(float(percent))


  


  Edit : FFMPEG console appearing and disappearing for a moment is shown in this video here in dropbox :
ffmpeg console appearing and disappearing video

  


• Sending video stream from NodeJS to python in real time [closed]

  17 juin 2021, par Tristan Delort

  I'm using a NodeJS server to catch a video stream through a WebRTC PeerConnection and I need to send it to a python script.

  


  I use NodeJS mainly because it's easy to use WebRTC in it and the package 'wrtc' supports RTCVideoSink and python's aiortc doesn't.

  


  I was thinking of using a named pipe with ffmpeg to stream the video stream but 3 questions arose :

  


  


  • Should I use python instead of NodeJS and completely avoid the stream through a named pipe part ? (This means there is a way to extract individual frames from a MediaStreamTrack in python)

    


  


  • If I stick with the "NodeJS - Python" approach, how do I send the stream from one script to the other ? Named pipe ? Unix domain sockets ? And with FFMpeg ?

    


  


  • Finally, for performance purpose I think that sending a stream and not each individual frames is better and simpler but is this true ?

    


  


  


  Thanks all !