Liquidsoap for Streaming

[WittyWidget]

Hi everyone,

Thanks to @toots and @gAlleb for the script.

I'm working with the script from this repository for my YouTube live streaming. I want to set up functionality to play requested tracks by viewers. If the requested track isn't already uploaded, I would like to upload it and play it immediately, similar to how AzuraCast works. Additionally, I want to integrate a chatbot to trigger the functions (skip, play, and add to queue). Here’s what I’m looking to do:

1. Skip: Skip the currently playing track. (http://host.com/skip like this?)
2. Play: Play a specific track immediately. If it’s not on the server, I will upload it and play it right away.
3. Add to Queue: Add a track to the queue to play after the current one finishes.

It would be much easier if there were a web UI similar to AzuraCast for managing these functions. I like AzuraCast, but I’m not sure if it can work with the ai-radio script while providing dynamic metadata display like Liquidsoap's ai-radio video canvas in real-time.

I don't need any AI functionality from the script, just the canvas with dynamic real-time updates.

Any help or advice would be greatly appreciated!

Thank you!

[gAlleb]

Hi @WittyWidget

1. First, you can definitely incorporate the script into Azuracast (with some tweaking via Edit Liquidsoap Configuration and adding necessary code there) and use its UI and queuing mechanisms to queue tracks.

• Also you can use its API to manipulate playlists (https://demo.azuracast.com/docs/api/).
  So it's definitely a good option if you want to save time. Usually "videocode" is pasted into the last box of Edit Liquidsoap Configuration.

Here is a very quick example which uses Azuracast default radio source, which can be pasted into the last box of the Edit Liquidsoap Configuration of your Azuracast station.

• It's compatible with LS 2.2.5 and uses cover manager function by @vitoyucepi.
• It doesn't use video.canvas but relies on real px values.
• In order for progressbar to work: radio_before_crossfade = radio should be inserted into the 3d box of Azuracast Liquidsoap Configuration (2.2.5 LS release).
• stream_key = "" is your youtube info and so on.

Example

# Store and retrieve file covers using metadata. This returns a set of
# getter/setter methods that can be used to store and retrieve cover art.
# Typical usage is to set cover art in a `on_metadata` handler and retrieve
# it in a `video.add_image` operator. See `video.add_cover` for an implementation
# example.
# @category Metadata
# @flag extra
# @param ~default Default cover file when no cover is available
# @param ~mime_types Recognized mime types and their corresponding file extensions.
def file.cover.manager(
  ~id=null(),
  ~mime_types=[
    ("image/gif", "gif"),
    ("image/jpeg", "jpeg"),
    ("image/jpg", "jpeg"),
    ("image/png", "png"),
    ("image/webp", "webp")
  ],
  ~default
) =
  id = string.id.default(id, default="cover")
  temp_dir = file.temp_dir("cover-manager")
  on_cleanup({file.rmdir(temp_dir)})

  last_cover_file = ref(default)
  current_cover_file = ref(default)

  def set_current_cover(filename) =
    last_cover = last_cover_file()
    last_cover_file := current_cover_file()
    current_cover_file := filename

    if
      last_cover != default
    then
      log.info(
        label=id,
        "Removing #{string.quote(last_cover)}"
      )
      file.remove(last_cover)
    end
  end

  def save_cover_to_file(cover) =
    mime = cover.mime
    extension = mime_types[mime]

    cover_file =
      if
        extension != ""
      then
        cover_file = file.temp("#{id}_", ".#{extension}")
        file.write(cover_file, data=cover)
        cover_file
      else
        null()
      end
    (cover_file, mime)
  end

  def extract_cover_from_metadata(_metadata) =
    filename = _metadata["filename"]
    log.info(
      label=id,
      "Extracting cover from #{string.quote(filename)}."
    )
    cover = metadata.cover(_metadata)
    let (cover_file, mime) =
      null.case(cover, {(null(), "")}, save_cover_to_file)

    next_cover =
      if
        cover == null()
      then
        log.important(
          label=id,
          "File #{string.quote(filename)} has no cover."
        )
        default
      elsif
        cover_file == null()
      then
        log.important(
          label=id,
          "File #{string.quote(filename)} has unknown mime type #{
            string.quote(mime)
          }."
        )
        default
      else
        cover_file = null.get(cover_file)
        log.important(
          label=id,
          "Cover for #{string.quote(filename)} saved to #{
            string.quote(cover_file)
          }."
        )
        cover_file
      end
    set_current_cover(next_cover)
  end

  current_cover_file.{set=extract_cover_from_metadata}
end
# Extract cover from the source's metadata and add it as a static image.
# @category Track / Video processing
# @flag extra
# @param ~id Force the value of the source ID.
# @param ~fallible Whether we are allowed to fail (in case the file is non-existent or invalid).
# @param ~width Scale to width
# @param ~height Scale to height
# @param ~x x position.
# @param ~y y position.
# @param ~default Default cover file when no cover is available
# @param ~mime_types Recognized mime types and their corresponding file extensions.
def video.add_cover(
  ~fallible=false,
  ~width=null(),
  ~height=null(),
  ~x=getter(0),
  ~y=getter(0),
  ~mime_types=[
    ("image/gif", "gif"),
    ("image/jpeg", "jpeg"),
    ("image/jpg", "jpeg"),
    ("image/png", "png"),
    ("image/webp", "webp")
  ],
  ~default,
  s
) =
  cover_file = file.cover.manager(mime_types=mime_types, default=default)
  s = source.on_metadata(s, cover_file.set)
  video.add_image(
     x=x, y=y, width=width, height=height, file=cover_file, s
  )
end

def string.of_int(~digits=0, n) =
  s = string(n)
  if
    string.length(s) >= digits
  then
    s
  else
    string.make(char_code=48, digits - string.length(s)) ^ s
  end
end

def remaining() =
  time = source.remaining(radio_before_crossfade)
  seconds = string.of_int(digits=2, int(time mod 60.))
  minutes = string.of_int(digits=2, int(time / 60.))
  "#{minutes}:#{seconds}"
end

# Custom Configuration (Specified in Station Profile)
base_dir = "/var/azuracast/stations/YOURSTATIONNAME"
 
stream_key = ""
background_dir = base_dir ^ "/videos"
 
font_file = base_dir ^ "/font.ttf"
font_size = 50
font_x = 340
font_y = 990
 
# progress bar
def source_percent(s)
  source.elapsed(s) / source.duration(s) * 100.0
end
# pre-calculate percent
px_percent = 1280.0/100.0
 
 
# We get NP info right from a radio source
 
nowplaying = ref("")
 
def update_nowplaying(m)
 
  if string.length(m["artist"]) <= 0 then
    if string.contains(substring=" - ", m["title"]) then
      let (a, t) = string.split.first(separator=" - ", m["title"])
           nowplaying := a ^ " - " ^ t
        end
  else
    nowplaying := m["artist"] ^ " - " ^ m["title"]
  end
end
 
radio.on_metadata(fun(m) -> thread.run(fast=false, {update_nowplaying(m)}))
 
videostream = playlist(background_dir)
 
videostream =
  video.add_rectangle(
    color=0xFF00FF,
    alpha=0.4,
    x=52,
    y=150,
    width=100,
    height=100,
    videostream
  )
 
 
videostream = video.add_rectangle(color=0x333333, alpha=0.7, x=925, y=50, width=355, height=55, videostream)

videostream =
  video.add_text(
    x=938, y=53, 
    size=43,
    color=0xFCB900,
    font=font_file,
    "Radio",
    videostream
  )
 
videostream = video.add_text(
      font=font_file, 
      size=font_size, 
      color=0xFFFFFF, 
      x=font_x, 
      y=font_y, 
      speed=0,
      {nowplaying()}, 
      videostream
)
videostream = video.add_image(x=52, y=52, width=100, height=100, file=base_dir ^ "/logo_500x500.png", videostream)
videostream = video.add_text(size=20, x=50, y=647,color=0xFFFFFF, {"Next in #{remaining()}"}, videostream)
videostream = video.add_rectangle(color=0xfcb900, x=0, y=673, height=10, width={int(px_percent*source_percent(radio_before_crossfade))}, videostream)
videostream = video.add_rectangle(color=0xfcb900, x=50, y=234, height=256, width=256, videostream)

videostream = source.mux.video(video=videostream, radio)

videostream = video.add_cover(x=53, y=237 , width=250, height=250, default=base_dir ^ "/logo_500x500.png",fallible=false, videostream)

# Output to YouTube
enc = %ffmpeg(
  format="flv", 
  %video(codec="libx264", pixel_format="yuv420p", b="300k", preset="superfast", r=25, g=50),
  %audio(
    codec="aac",
    samplerate=44100,
    channels=2,
    b="128k",
    profile="aac_low"
  )
)
output.url(url=stream_key, enc, fallible=true, videostream)

2. If you choose to make everything yourself I encourage you to read LS docs and the book (https://www.liquidsoap.info/book/book.pdf) which has most of the answers you need to come up with a solution that suits you.

3. I want to integrate a chatbot to trigger the functions (skip, play, and add to queue)

• If you want harbor functions like this - it's quite easy to implement. Most of them are described in a book as well. Skip:

# if your source is radio:
# Skip - will be available via http://localhost:8007/skip

def skipper(_)
radio.skip()
http.response(data="The current song was skipped!")
end
harbor.http.register.simple(port=8007, "/skip", skipper)

• Add to Queue: Add a track to the queue to play after the current one finishes. Create a queue and push files via http://localhost:8007/play. Tracks then can be feeded via http://localhost:8007/play?file=/path/to/track

queue = request.queue()
radio = fallback(track_sensitive=true, [queue, radio])

def play(request)
fname = request.query["file"]
log.important("Requested file #{fname}.")

if file.exists(fname) then
log.important("File Request Accepted.")
queue.push.uri("#{fname}")
http.response(content_type="text/html", status_code=200, data="<p>Request pushed.</p>")
else
log.important("File Request Rejected File Not Found.")
http.response(content_type="text/html", status_code=404, data="<p>Invalid file: #{fname}</p>")
end
end

harbor.http.register.simple("/play", play, port=8007, method="GET")

• Play a specific track immediately.

# Play: Play a specific track immediately. Create an `interupting_queue` 
# and push files via http://localhost:8007/play_now

interrupting_queue = request.queue()
radio = fallback(track_sensitive=false, [interrupting_queue, radio])

def play(request)
fname = request.query["file"]
log.important("Requested file #{fname}.")

if file.exists(fname) then
log.important("File Request Accepted.")
interrupting_queue.push.uri("#{fname}")
http.response(content_type="text/html", status_code=200, data="<p>Request pushed.</p>")
else
log.important("File Request Rejected File Not Found.")
http.response(content_type="text/html", status_code=404, data="<p>Invalid file: #{fname}</p>")
end
end

harbor.http.register.simple("/play_now", play, port=8007, method="GET")

• Next you can hide 8007 port behind proxy to get the desired url: e.g. http://host.com/play?file=...

And so on.

---

• Also you can read about video streaming with LS in the blog by @Mikulskii - https://mikulski.rocks/blog/. He streams to youtube: https://www.youtube.com/c/Mikulski_Youtube

• And here is The Great Deer Radio https://github.com/StarveTheEgo/deer-radio (@StarveTheEgo): https://www.youtube.com/watch?v=R4qlvDjSPUA&pp=ygUKZGVlciByYWRpbw%3D%3D

• Nightride.fm: https://nightride.fm/video

So everything is relatively easy if you you are eager to invest some time and learn something new.

[WittyWidget]

Thank you very much, @gAlleb, for the answer. I just tested the harbor functions (skip, play, and play_now). They are working, but not quite like in Azuracast. There's a delay of 10-15 seconds for skip and 7 seconds for play_now, whereas in Azuracast, the skip/play happens immediately.

Does Azuracast have any separate configuration for that?

Thank you again for helping the community.

[gAlleb]

The skip function is the same. You can see it from the AC config files. The latency can be caused by a huge number of cases - mainly the latency between you and icecast instance.

Not sure about what play you are talking about. I'm not aware of play_now in AC. You can play any file from AC media folder but it's not connected to the stream.

Maybe there is a difference between skip implementation via telnet (AC) and a harbor function but it should be minimal.

You can try with a telnet command:

settings.server.telnet := true
settings.server.telnet.port := 1234
s = playlist(id="music", "~/Music")

server.register(
usage="skip_test",
description="Skip the current song.",
"skip_test",
fun(_) -> begin source.skip(s) "Done!" end
)
output(s)

Then connect to telnet localhost 1234 and try to skip_test or probably there is a native implementation of skip is already in telnet.

[WittyWidget]

Dear @gAlleb,

1. The harbor functions are tested in standalone Liquidsoap, specifically in this ai-radio template which causes the delay of skip, and play_now.

2. Added the below code as per your instructions in Azuracast "Edit Liquidsoap configuration" option.

Liquidsoap code for AzuraCast - RTMP Video Stream

# Store and retrieve file covers using metadata. This returns a set of
# getter/setter methods that can be used to store and retrieve cover art.
# Typical usage is to set cover art in a `on_metadata` handler and retrieve
# it in a `video.add_image` operator. See `video.add_cover` for an implementation
# example.
# @category Metadata
# @flag extra
# @param ~default Default cover file when no cover is available
# @param ~mime_types Recognized mime types and their corresponding file extensions.
def file.cover.manager(
  ~id=null(),
  ~mime_types=[
    ("image/gif", "gif"),
    ("image/jpeg", "jpeg"),
    ("image/jpg", "jpeg"),
    ("image/png", "png"),
    ("image/webp", "webp")
  ],
  ~default
) =
  id = string.id.default(id, default="cover")
  temp_dir = file.temp_dir("cover-manager")
  on_cleanup({file.rmdir(temp_dir)})

  last_cover_file = ref(default)
  current_cover_file = ref(default)

  def set_current_cover(filename) =
    last_cover = last_cover_file()
    last_cover_file := current_cover_file()
    current_cover_file := filename

    if
      last_cover != default
    then
      log.info(
        label=id,
        "Removing #{string.quote(last_cover)}"
      )
      file.remove(last_cover)
    end
  end

  def save_cover_to_file(cover) =
    mime = cover.mime
    extension = mime_types[mime]

    cover_file =
      if
        extension != ""
      then
        cover_file = file.temp("#{id}_", ".#{extension}")
        file.write(cover_file, data=cover)
        cover_file
      else
        null()
      end
    (cover_file, mime)
  end

  def extract_cover_from_metadata(_metadata) =
    filename = _metadata["filename"]
    log.info(
      label=id,
      "Extracting cover from #{string.quote(filename)}."
    )
    cover = metadata.cover(_metadata)
    let (cover_file, mime) =
      null.case(cover, {(null(), "")}, save_cover_to_file)

    next_cover =
      if
        cover == null()
      then
        log.important(
          label=id,
          "File #{string.quote(filename)} has no cover."
        )
        default
      elsif
        cover_file == null()
      then
        log.important(
          label=id,
          "File #{string.quote(filename)} has unknown mime type #{
            string.quote(mime)
          }."
        )
        default
      else
        cover_file = null.get(cover_file)
        log.important(
          label=id,
          "Cover for #{string.quote(filename)} saved to #{
            string.quote(cover_file)
          }."
        )
        cover_file
      end
    set_current_cover(next_cover)
  end

  current_cover_file.{set=extract_cover_from_metadata}
end
# Extract cover from the source's metadata and add it as a static image.
# @category Track / Video processing
# @flag extra
# @param ~id Force the value of the source ID.
# @param ~fallible Whether we are allowed to fail (in case the file is non-existent or invalid).
# @param ~width Scale to width
# @param ~height Scale to height
# @param ~x x position.
# @param ~y y position.
# @param ~default Default cover file when no cover is available
# @param ~mime_types Recognized mime types and their corresponding file extensions.
def video.add_cover(
  ~fallible=false,
  ~width=null(),
  ~height=null(),
  ~x=getter(0),
  ~y=getter(0),
  ~mime_types=[
    ("image/gif", "gif"),
    ("image/jpeg", "jpeg"),
    ("image/jpg", "jpeg"),
    ("image/png", "png"),
    ("image/webp", "webp")
  ],
  ~default,
  s
) =
  cover_file = file.cover.manager(mime_types=mime_types, default=default)
  s = source.on_metadata(s, cover_file.set)
  video.add_image(
     x=x, y=y, width=width, height=height, file=cover_file, s
  )
end

def string.of_int(~digits=0, n) =
  s = string(n)
  if
    string.length(s) >= digits
  then
    s
  else
    string.make(char_code=48, digits - string.length(s)) ^ s
  end
end

def remaining() =
  time = source.remaining(radio_before_crossfade)
  seconds = string.of_int(digits=2, int(time mod 60.))
  minutes = string.of_int(digits=2, int(time / 60.))
  "#{minutes}:#{seconds}"
end

# Custom Configuration (Specified in Station Profile)
base_dir = "/var/azuracast/stations/YOURSTATIONNAME"
 
stream_key = ""
background_dir = base_dir ^ "/videos"
 
font_file = base_dir ^ "/font.ttf"
font_size = 50
font_x = 340
font_y = 990
 
# progress bar
def source_percent(s)
  source.elapsed(s) / source.duration(s) * 100.0
end
# pre-calculate percent
px_percent = 1280.0/100.0
 
 
# We get NP info right from a radio source
 
nowplaying = ref("")
 
def update_nowplaying(m)
 
  if string.length(m["artist"]) <= 0 then
    if string.contains(substring=" - ", m["title"]) then
      let (a, t) = string.split.first(separator=" - ", m["title"])
           nowplaying := a ^ " - " ^ t
        end
  else
    nowplaying := m["artist"] ^ " - " ^ m["title"]
  end
end
 
radio.on_metadata(fun(m) -> thread.run(fast=false, {update_nowplaying(m)}))
 
videostream = playlist(background_dir)
 
videostream =
  video.add_rectangle(
    color=0xFF00FF,
    alpha=0.4,
    x=52,
    y=150,
    width=100,
    height=100,
    videostream
  )
 
 
videostream = video.add_rectangle(color=0x333333, alpha=0.7, x=925, y=50, width=355, height=55, videostream)

videostream =
  video.add_text(
    x=938, y=53, 
    size=43,
    color=0xFCB900,
    font=font_file,
    "Radio",
    videostream
  )
 
videostream = video.add_text(
      font=font_file, 
      size=font_size, 
      color=0xFFFFFF, 
      x=font_x, 
      y=font_y, 
      speed=0,
      {nowplaying()}, 
      videostream
)
videostream = video.add_image(x=52, y=52, width=100, height=100, file=base_dir ^ "/logo_500x500.png", videostream)
videostream = video.add_text(size=20, x=50, y=647,color=0xFFFFFF, {"Next in #{remaining()}"}, videostream)
videostream = video.add_rectangle(color=0xfcb900, x=0, y=673, height=10, width={int(px_percent*source_percent(radio_before_crossfade))}, videostream)
videostream = video.add_rectangle(color=0xfcb900, x=50, y=234, height=256, width=256, videostream)

videostream = source.mux.video(video=videostream, radio)

videostream = video.add_cover(x=53, y=237 , width=250, height=250, default=base_dir ^ "/logo_500x500.png",fallible=false, videostream)

# Output to YouTube
enc = %ffmpeg(
  format="flv", 
  %video(codec="libx264", pixel_format="yuv420p", b="300k", preset="superfast", r=25, g=50),
  %audio(
    codec="aac",
    samplerate=44100,
    channels=2,
    b="128k",
    profile="aac_low"
  )
)
output.url(url=stream_key, enc, fallible=true, videostream)

Now, I have tested the built-in 'Skip' button available on the AzuraCast web interface, which instantly triggers the current track to skip. However, when using standalone Liquidsoap (ai-radio), there is a delay of 10-15 seconds with the harbor/telnet functions.

My apologies, I'm trying to understand the code, but I'm getting stuck.

Do you have a full template (ai-radio) canvas for AzuraCast configuration? The provided template only displays the album cover and progress bar. The AzuraCast method seems easier.

I would be grateful if you could upload the full template for AzuraCast Users in your repository.

Many thanks!

[gAlleb]

The provided template only displays the album cover and progress bar. The AzuraCast method seems easier.

It displays other things too

1. NP info:

videostream = video.add_text(
      font=font_file, 
      size=font_size, 
      color=0xFFFFFF, 
      x=font_x, 
      y=font_y, 
      speed=0,
      {nowplaying()}, 
      videostream
)

2. Remaining time.

videostream = video.add_text(size=20, x=50, y=647,color=0xFFFFFF, {"Next in #{remaining()}"}, videostream)

3. ...

It's a full template. What can be fuller? Paste radio_before_crossfade = radio into the 3d box of Azuracast Liquidsoap Configuration^ other code at the end. Edit width/height parameters and colors to suit your needs.