mp4ff

Package mp4ff implements MP4 media file parser and writer for AVC video, AAC audio and stpp/wvtt subtitles. Focused on fragmented files as used for streaming in DASH, MSS and HLS fMP4.

Library

The library has functions for parsing (called Decode) and writing (Encode) in the package mp4ff/mp4. It also contains codec specific parsing of of AVC/H.264 including complete parsing of SPS and PPS in the package mp4ff.avc.

Traditional multiplexed non-fragmented mp4 files can also be parsed and decoded, see examples/segment.

The focus is, however, on non-multiplexed single-track fragmented mp4 files as used in DASH, HLS, and CMAF.

The top level structure for both non-fragmented and fragmented mp4 files is mp4.File.

In the non-fragmented files, the members Ftyp, Moov, and Mdat are used. A fragmented mp4.File file can be a single init segment, one or more media segments, or a a combination of both like a CMAF track which renders into a playable one-track asset.

The following high-level structures are used:

• InitSegment contains an ftyp and moov box and provides the metadata for a fragmented files. It corresponds to a CMAF header
• MediaSegment starts with an optional styp box and contains on or more Fragments
• Fragment is an mp4 fragment with exactly one moof box followed by a mdat box where the latter contains the media data. It is limited to have exactly one trun box.

The typical child boxes are exported so that one can write paths such as

fragment.Moof.Traf.Trun

to access the (only) trun box in a fragment.

The codec currently supported are AVC/H.264 and AAC.

Usage for creating new fragmented files

A typical use case is to produce an init segment and followed by a series of media segments.

The first step is to create an init segment. This is done in two steps and can be seen in `examples/initcreator:

1. A call to CreateEmptyInitSegment(timecale, mediatype, language)
2. Fill in the SampleDescriptor based on video parameter sets, or audio codec information

The second step is to start producing media segments. They should use the timescale that was set when creating the init segments. The timescales should be chosen so that the sample durations have exact values.

A media segment contains one or more fragments, where each fragment has a moof and an mdat box. If all samples are available before the segment is to be used, one can use use a single fragment in each segment. Example code for this can be found in examples/segmenter.

The high-level code to do that is to first create a slice of FullSample with the data needed. All times are in the track timescale set when creating the init segment and coded in the mdhd box.

mp4.FullSample{
	Sample: mp4.Sample{
        Flags uint32 // Flag sync sample etc
        Dur   uint32 // Sample duration in mdhd timescale
        Size  uint32 // Size of sample data
        Cto   int32  // Signed composition time offset
	},
    DecodeTime uint64 // Absolute decode time (offset + accumulated sample Dur)
    Data       []byte // Sample data
}

The mp4.Sample part is what will be written into the trun box. DecodeTime is the media timeline accumulated time. The value of the first samples of a fragment, will be set as the BaseMediaDecodeTime in the tfdt box when calling Encode on the fragment. Once a number of such samples are available, then can be added to a media segment

seg := mp4.NewMediaSegment()
frag := mp4.CreateFragment(uint32(segNr), mp4.DefaultTrakID)
seg.AddFragment(frag)
for _, sample := range samples {
	frag.AddSample(sample)
}

This segment can finally be output to a w io.Writer as

err := seg.Encode(w)

Direct changes of attributes

Many attributes are public and can therefore be changed in an uncontrolled way. The advantage of this is that it is possible to write code that can manipulate boxes in many different ways. However, some attributes and linkd can become broken.

As an example, container boxes such as TrafBox have a method AddChild which adds a box to its slice of children boxes Children, but also sets a specific member reference such as Tfdt to point to that box. If Children is manipulated directly, that link will not be valid.

Automatic settings of values on Fragment.Encode

It is possible to optimize the TrunBox size by writing default values in TfhdBox. To do this, one must analyze if, for example, the duration of all samples is the same. This is done by the method TrafBox.OptimizeTfhdTrun which changes its children Tfhdand Trun. Since this will change the size of boxes, it is important that this function is called before Encode on any parent box of Traf. In particular, this method is called automatically when running Fragment.Encode.

Another value which is automatically set by Moof.Encode is MoofBox.Traf.Trun.DataOffset. This value is the address of the first media sample relative to the start of the MoofBox start. It therefore depends on the size of MoofBox and is unknown until all values are in place so that it can be calculated. It is set to MoofBox.Size()+8.

Command Line Tools

Some simple command line tools are available in cmd.

1. mp4ff-pslister extracts and displays pps and sps for AVC in an mp4 file.
2. mp4ff-info prints a tree of the box hierarchy of an mp4 file with information of the boxes. The level of detail can be increased with the option -l, like -l all:1 for all boxes or -l trun:1,stss:1 for specific boxes.

You can install them by going to their respective directory and run go install.

Example code

Example code is available in the examples directory. These are

1. initcreator which creates typical init segments (ftyp + moov) for video and audio
2. resegmenter which reads a segmented file (CMAF track) and resegments it with other segment durations.
3. segmenter which takes a progressive mp4 file and creates init and media segments from it.

Stability

The APIs should be fairly stable, but minor non-backwards-compatible tweaks may happen until version 1.

LICENSE

MIT, see LICENSE.md.

Some code in pkg/mp4, comes from or is based on https://github.com/jfbus/mp4 which has Copyright (c) 2015 Jean-François Bustarret.

Some code in pkg/bits comes from or is based on https://github.com/tcnksm/go-casper/tree/master/internal/bits Copyright (c) 2017 Taichi Nakashima.

Versions

See Versions.md.