Prologue 
--------

If you were waiting for the next installment of the TTY office series, 
I got a bit sidetracked.

A few weeks ago, Rogers Communications shut down six radio stations 
across Canada. Two were in Vancouver. One was a 24-hour news station 
(CKWX) and the other was a sports talk station (CISL, better known as 
"Sportsnet 650"). These were AM stations. I tended to listen online 
during the day and over the air at night, due to my distance from the 
transmitters. After they shut down, I decided to set up my own 
streaming sports talk station.

My setup makes use of ezstream and icecast2, both of which are 
available in the Debian repositories. 

Here's a flowchart to outline how these pieces of software function 
together:

 ezstream                  icecast2                  various
(streaming     ----->     (streaming     ----->     receiving
 client)                    server)                  clients

The streaming client sends the stream to the streaming server, which 
then relays it to receiving clients. The streaming client and server 
may be on different machines. Mine are. I have ezstream installed on 
my laptop at home. It sends the stream to icecast2 on my vps in 
Toronto. For clients, I use Transistor to listen on my Android phone 
and gst123 to listen on my laptop.


Configuration
-------------

Here are my configuration files for ezstream and icecast2:

I called my ezstream configuration file ez-config.xml. It can be 
placed in any convenient directory. I've used "password_one" as the 
password in this file so you can connect it to the correct password in 
the icecast.xml file. If you have icecast on the same device as 
ezstream, substitute 127.0.0.1 for your.server.com in this file.

<ezstream>

  <servers>
    <server>
      <hostname>your.server.com</hostname>
      <port>8000</port>
      <password>password_one</password>
    </server>
  </servers>

  <streams>
    <stream>
      <mountpoint>/stream.ogg</mountpoint>
      <format>Ogg</format>
      <stream_name>AM 750 CSPO</stream_name>
       <stream_description>BC's Sports Talk Connection!</stream_description>
    </stream>
  </streams>

  <intakes>
    <intake>
      <filename>stdin</filename>
    </intake>
  </intakes>

</ezstream>


The icecast.xml configuration file is at /etc/icecast2/icecast.xml. 
It's a long file and you only need to make a few changes. Here are the 
changed sections of mine, separated by ellipses. In this file, 
"password_one" should be the same password you used in the ezstream 
configuration file. You use "password_three" to log in as 
administrator in icecast's web portal, which you'll find by visiting 
"your.server.com:8000" with your web browser. Unless you're dealing 
with multiple icecast instances, you'll probably never use 
"password_two". These particular passwords are used for illustrative 
purposes. Change them to something more difficult to guess!

    <authentication>
        <!-- Sources log in with username 'source' -->
        <source-password>password_one</source-password>
        <!-- Relays log in with username 'relay' -->
        <relay-password>password_two</relay-password>

        <!-- Admin logs in with the username given below -->
        <admin-user>ArthurCarlson</admin-user>
        <admin-password>password_three</admin-password>
    </authentication>

...

    <hostname>your.server.com</hostname>
    <!-- You may have multiple <listen-socket> elements -->
    <listen-socket>
        <port>8000</port>
        <!--<bind-address>170.75.161.211</bind-address>-->
        <!--<shoutcast-mount>/stream</shoutcast-mount>-->
    </listen-socket>

...

    <mount type="normal">
        <mount-name>/stream.ogg</mount-name>
        <max-listener-duration>14400</max-listener-duration>
        <fallback-mount>/static.ogg</fallback-mount>
        <fallback-override>1</fallback-override>
        <fallback-when-full>1</fallback-when-full>
    </mount>


The fallback mount, static.ogg, is actually just a file placed in the 
/usr/share/icecast2/web/ directory. It plays a one-minute recording of 
radio static when the station is offline (when I'm not relaying any 
streams through ezstream). Each time icecast2 concludes playing the 
file, it will check to see if I'm streaming again and will attempt to 
switch listeners over to the main stream ("stream.ogg"). The 
switch-over only seems to work with well-behaved clients. For the 
fallback stream, you can use any file you like, but make it short, 
since it doesn't appear to try to hand clients back to the main stream 
until the file finishes playing.

If you read up further about icecast2, you'll find that you can 
actually set up multiple "mounts" (to which clients can send streams) 
and prioritize among them by setting up chains of fallbacks. You can 
also use "liquidsoap" to do this kind of thing in a more professional 
way, but I haven't tried it.


To start icecast2: 

$ systemctl start icecast2.service

To autostart icecast2 each time you start the server:

$ systemctl enable icecast2.service

You will listen to your station at: your.server.com:8000/stream.ogg

But first, you have to send it a stream. 


Streaming "Recipes"
-------------------

Okay, here comes the fun part! Here are some one-line recipes you can 
use to stream from various sources to your station. In several of 
these recipes, I use yt-dlp to stream from YouTube and pipe the output 
through ffmpeg and ezstream. Since I'm the only person using this 
station, I'm not really concerned about copyright.

It's best to download the yt-dlp binary directly from their releases 
on Github (https://github.com/yt-dlp/yt-dlp). Place it in a directory 
in your path. I've renamed it from "yt-dlp_linux" to "yt-dlp". If you 
leave it as yt-dlp_linux, just substitute that name in the commands 
below. Once installed, you can update yt-dlp to the latest version by 
typing "yt-dlp -U" without the quotation marks. 

You'll also need ffmpeg. It's in the Debian repos.

Here are the recipes:

# Send a local file to your station at 64k (stereo):
ffmpeg -i path/to/file.mp3 -b:a 64k -f opus - | ezstream -c /path/to/ez-config.xml

# Relay another audio or video stream to your station as 16k mono audio:
ffmpeg -i https://rogers-hls.leanstream.co/rogers/tor590.stream/playlist.m3u8 -b:a 16k -f opus -ac 1 - | ezstream -c /path/to/ez-config.xml

# Relay a YouTube stream to your station at 64k (stereo):
yt-dlp -f wa* https://www.youtube.com/watch?v=6gtQck24Nf8 -o - | ffmpeg -i pipe: -b:a 64k -f opus - | ezstream -c /path/to/ez-config.xml

# Relay and record a YouTube stream simultaneously in 16k mono:
yt-dlp -f wa* https://www.youtube.com/watch?v=DbO0t3c5ukk -o - | ffmpeg -i pipe: -b:a 16k -f opus -ac 1 - | tee testfile.ogg | ezstream -c /path/to/ez-config.xml

# Relay a YouTube livestream to your station at 16k mono (with some 
# sound modifications to make it sound like an AM station). See the 
# switch explanations below.
timeout 120m yt-dlp --wait-for-video 30-60 -f wa* https://www.youtube.com/@SekeresPrice/live -o - | ffmpeg -i pipe: -b:a 16k -f opus -af "asubboost=boost=2.0:cutoff=60,equalizer=f=1500:width_type=h:width=200:g=1.2,lowpass=f=6000,volume=0.8" -ac 1 - | ezstream -c /path/to/ez-config.xml

What do the various switches and commands mean?

ffmpeg
 -i                             input file (or stream or pipe)
 -b:a 64k                       bitrate audio (set to 64k)
 -f opus                        encode audio as opus (the format used inside ogg containers)
 -ac 1                          audio channel 1 (mono rather than stereo)
 -                              output to stdout, so you can pipe it to the next application
 -af                            audio filter

yt-dlp
 --wait-for-video 30-60         wait for the livestream to start. Wait at least 30 seconds and at most 60 seconds between retries.
 -f wa*                         get the worst quality audio-only (preferred) or audio-video stream from YouTube
 -o -                           output to stdout
 
ezstream
 -c                             location of configuration file

tee                             adds a "tee" into your pipe: 

                                         testfile.ogg
                                               ^
                                yt-dlp > ffmpeg > ezstream  

timeout 120m                    kill the command if it's still running after 120 minutes

Here's one final recipe. It's a bit more complicated to set up. This 
one relays the sound from your device (any application playing sound 
from any source) to ezstream and on to your icecast station. So you 
can just play something using an audio player, your browser, etc., and 
you'll hear the sound on your icecast station.[1]

ffmpeg -f pulse -i <audio sink> -b:a 16k -f opus -ac 1 - | ezstream -c /path/to/ez-config.xml

To make this command work, you need to identify your default audio 
sink and substitute it into the command. Here are the steps:

1. Identify your default audio sink (output device):

$ pactl get-default-sink

For illustrative purposes, here's the output of that command on my 
device: 

alsa_output.pci-0000_00_1b.0.analog-stereo

2. Identify the sink source that you're going to add to the recipe:

$ pactl list sinks | grep <YOUR DEFAULT SINK FROM PREVIOUS COMMAND OUTPUT>

So for me, the command is:

$ pactl list sinks | grep alsa_output.pci-0000_00_1b.0.analog-stereo

My device returns this output:
	Name: alsa_output.pci-0000_00_1b.0.analog-stereo
	Monitor Source: alsa_output.pci-0000_00_1b.0.analog-stereo.monitor
		node.name = "alsa_output.pci-0000_00_1b.0.analog-stereo"

Look for a line ending with the word "monitor" and add it into the 
recipe. Here's the final recipe using the correct sink source for my 
device:

ffmpeg -f pulse -i alsa_output.pci-0000_00_1b.0.analog-stereo.monitor -b:a 16k -f opus -ac 1 - | ezstream -c /path/to/ez-config.xml

If you want to automate your station, put the one-line recipes into 
bash scripts and set them to run automatically with cronjobs.

If there are errors in this post, or you have suggestions, let me know!

visiblink (at) gmx (dot) com

___________________

[1] This information is from 
    https://linuxconfig.org/recording-browser-audio-on-linux-with-ffmpeg