FFmpeg Support

Since the 2.0.x release cycle, liquidsoap integrates a tight support of ffmpeg. This includes:

Ffmpeg support includes 3 types of content:

  • Internal content, that is content available to all liquidsoap operators: PCM audio and YUV420p video
  • Raw content, that is decoded content but stored as ffmpeg internal frame. This type of content is only available to ffmpeg filters and raw encoders. It can be used to avoid data copies back and forth between liquidsoap and ffmpeg.
  • Copy content, that is encoded content stored as ffmpeg internal packets. This type of content is only available to ffmpeg copy encoder and bitstream filters and requires a fairly good understanding of media codecs and containers. Copy contents can be used to avoid transcoding and pass encoded data end-to-end inside liquidsoap scripts.

Enabling ffmpeg support

FFmpeg support is available via the external ocaml-ffmpeg binding package. If you are using any binary asset from our release pages or via docker, this should already be included.

If you are installing via opam, installing the ffmpeg package should do the trick:

% opam install ffmpeg

Decoders

For the most part, you should never have to worry about the ffmpeg decoder. When enabled, it should be the preferred decoder for all supported media. When using raw or copied content, the decoder is able to produce the required content without the need of any intervention on the user part.

Should you need to tweak it, here are a couple of pointers:

The settings.decoder.decoders settings controls which decoders are to be used when trying to decode media files. You can use it to restrict which decoders are being used, for instance making sure only the ffmpeg decoder is used:

settings.decoder.decoders.set(["FFMPEG"])

Priority for the decoder is set via:

settings.decoder.priorities.ffmpeg.set(10)

You can use this setting to adjust wether or not the ffmpeg decoder should be tried first when decoding media files, in particular in conjunction with the other settings.decoder.priorities.* settings.

For each type of media codec, the settings.decoder.ffmpeg.codecs.* settings can be used to tell ffmpeg which decoder to use to decode this type of content (there could more than one decoder for a given codec).

For instance, for the aac codec:

  • settings.decoder.ffmpeg.codecs.aac.available() returns the list of available decoders, typically ["aac", "aac_fixed"].
  • settings.decoder.ffmpeg.codecs.aac.set can be used to choose which decoder should be used, typically: settings.decoder.ffmpeg.codecs.aac.set("aac")

When debugging issues with ffmpeg, it can be useful to increase the log verbosity.

settings.ffmpeg.log.verbosity.set("warning")

This settings sets the verbosity of ffmpeg logs. Possible values, from less verbose to more verbose are: "quiet", "panic", "fatal", "error", "warning", "info", "verbose" or "debug"

Please note that, due to a technical limitation, we are not yet able to route ffmpeg logs through the liquidsoap logging facilities, which means that ffmpeg logs are currently only printed to the process’s standard output and that the settings.ffmpeg.log.level is currently not used.

Encoders

See detailed ffmpeg encoders article.

Filters

See detailed ffmpeg filters article.

Bitstream filters

FFmpeg bitstream filters are filters that modify the binary content of encoded data. They can be used to adjust certain aspects of media codecs and containers to make them fit some specific use, for instance a rtmp/flv output etc. They are particularly important when dealing with live switches of encoded content (see Examples section).

The list of all bitstream filters is documented on FFmpeg online doc and our extra API reference. Here’s one such filter:

% liquidsoap -h ffmpeg.filter.bitstream.h264_mp4toannexb

FFmpeg h264_mp4toannexb bitstream filter. See ffmpeg documentation for more
details.

Type: (?id : string?, source(audio='a,
 video=ffmpeg.video.copy('b), midi=none)) ->
source(audio='a, video=ffmpeg.video.copy('b), midi=none)

Category: Source / FFmpeg Filter
Flag: extra

Parameters:

 * id : string? (default: null)
     Force the value of the source ID.

 * (unlabeled) : source(audio='a, video=ffmpeg.video.copy('b), midi=none)

Methods:
...

Please consult the FFmpeg documentation for more details about that each filter do and why/how to use them.

Encoded data tweaks

Manipulating encoded content is powerful but can sometimes require some specific knowledge of internals aspects of media codecs and containers. This section lists some specific cases.

Relaxed copy content compatibility check

By default, liquidsoap keeps track of the content passed in a stream containing ffmpeg encoded content (ffmpeg.copy) and only allows file and stream decoders to return strictly compatible content, e.g. same video resolution or audio samplerate.

Some containers such as mp4, however, do allow stream where video resolution or audio samplerate changes between tracks. In this case, you can relax those compatibility checks using the following setting:

settings.ffmpeg.content.copy.relaxed_compatibility_check.set(true)

This is a global setting for now and could be refined per-stream in the future if the needs arises.

Shared encoders

liquisoap provides operators to encode data using %ffmpeg and re-use it accross output. This is called inline encoding. Here’s an example:

audio_source = single(audio_url)
video_source = single(image)

stream = mux_video(video=video_source, audio_source)

stream = ffmpeg.encode.audio_video(
    %ffmpeg(
        %audio(codec="aac", b="128k"),
        %video(codec="libx264", b="4000k")
    ),
    stream
)

flv = %ffmpeg(
    format="flv",
    %audio.copy,
    %video.copy,
)

# Send to one youtube output:
output.youtube.live.rtmp(
    encoder = flv,
    stream,
    ...
)

mpegts = %ffmpeg(
    format="mpegts",
    %audio.copy,
    %video.copy,
)

# And to a hls one:
output.file.hls(
  ["mpegts", mpegts],
  stream,
  ...
)

Working with encoded data, however, requires a bit of knowledge of ffmpeg internal and media codecs and containers. Here, for instance, this stream will have issues because the flv format requires global data, something that in ffmpeg terms is called extradata.

When working with a single encoder such as:

%ffmpeg(
  format="flv",
  %audio(codec="aac", b="128k"),
  %video(codec="libx264", b="4000k")
)

We are aware when initializing the encoders that it is aimed for a flv container so the code implicitely enables the global header for each encoder.

However, when encoding inline, we do not know at the time of encoding the container that will be used to encapsulate the stream, even worst, it can be used potentially with different containers with different requirements!

In our case here, you have two ways to solve the issue:

If you know that all the containers will be okay with global header, you can enable the corresponding flag in the encoder:

stream = ffmpeg.encode.audio_video(
    %ffmpeg(
        %audio(codec="aac", b="128k", flags="+global_header"),
        %video(codec="libx264", b="4000k", flags="+global_header")
    ),
    stream
)

However, it is also possible that one stream needs global header but not the other one, which is the case here with mpegts. In this case, you can use the bitstream filter ffmpeg.filter.bitstream.extract_extradata to extract global data to only one stream:

audio_source = single(audio_url)
video_source = single(image)

stream = mux_video(video=video_source, audio_source)

stream = ffmpeg.encode.audio_video(
    %ffmpeg(
        %audio(codec="aac", b="128k"),
        %video(codec="libx264", b="4000k")
    ),
    stream
)

flv = %ffmpeg(
    format="flv",
    %audio.copy,
    %video.copy,
)

flv_stream = ffmpeg.filter.bitstream.extract_extradata(stream)

# Send to one youtube output:
output.youtube.live.rtmp(
    encoder = flv,
    flv_stream,
    ...
)

mpegts = %ffmpeg(
    format="mpegts",
    %audio.copy,
    %video.copy,
)

# And to a hls one:
output.file.hls(
  ["mpegts", mpegts],
  stream,
  ...
)

Examples

See detailed ffmpeg cookbook article.