Membrane H26x Plugin

Hex.pmAPI DocsCircleCI

Membrane H.264 and H.265 parsers. It is a pair of Membrane elements responsible for parsing the incoming H.264 and H.265 streams. The parsing is done as a sequence of the following steps:

The output of the element is the incoming binary payload, enriched with the metadata describing the division of the payload into access units.

It is part of Membrane Multimedia Framework.

Installation

The package can be installed by adding membrane_h26x_plugin to your list of dependencies in mix.exs:

def deps do
[
{:membrane_h26x_plugin, "~> 0.11.0"}
]
end

Usage

The following pipeline takes H264 file, parses it, and then decodes it to the raw video.

defmodule Decoding.Pipeline do
use Membrane.Pipeline
alias Membrane.{File, H264}
@impl true
def handle_init(_ctx, _opts) do
spec =
child(:source, %File.Source{location: "test/fixtures/input-10-720p-main.h264"})
|> child(:parser, H264.Parser)
|> child(:decoder, H264.FFmpeg.Decoder)
|> child(:sink, %File.Sink{location: "output.raw"})
{[spec: spec], nil}
end
@impl true
def handle_element_end_of_stream(:sink, _ctx_, state) do
{[terminate: :normal], state}
end
end

Parsing without Membrane

The parsing logic itself lives in Membrane.H26x.ParsingEngine and can be used standalone, without spawning any Membrane component. The following snippet parses an H264 file and returns its access units:

alias Membrane.H26x.ParsingEngine
engine =
ParsingEngine.new(%{
codec: :h264,
input_stream_structure: :annexb,
input_alignment: :bytestream
})
{access_units, engine} = ParsingEngine.push(engine, File.read!("video.h264"))
{last_access_units, _engine} = ParsingEngine.flush(engine)
access_units = access_units ++ last_access_units

Each access unit is a list of Membrane.H26x.NALu structs, carrying the NALus' types, payloads and parsed fields. The payload can also be fed in arbitrarily split chunks with repeated ParsingEngine.push/3 calls - each call returns the access units completed so far, and ParsingEngine.flush/1 drains whatever remains buffered.

For length-prefixed streams (e.g. tracks demuxed from an MP4 container), pass the Decoder Configuration Record binary as the input stream structure - the NALu length size is read from it and the parameter sets it carries are parsed along with the first pushed payload:

engine =
ParsingEngine.new(%{
codec: :h264,
input_stream_structure: {:avc1, dcr},
input_alignment: :au
})

Copyright 2022, Software Mansion

Software Mansion

Licensed under the Apache License, Version 2.0