This document defines the Matroska audiovisual container, including definitions of its structural elements, as well as its terminology, vocabulary, and application.¶
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task
Force (IETF). Note that other groups may also distribute working
documents as Internet-Drafts. The list of current Internet-Drafts is
at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 29 April 2020.¶
Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with
respect to this document. Code Components extracted from this
document must include Simplified BSD License text as described in
Section 4.e of the Trust Legal Provisions and are provided without
warranty as described in the Simplified BSD License.¶
Matroska aims to become THE standard of multimedia container formats. It was derived from a project called MCF, but differentiates from it significantly because it is based on EBML (Extensible Binary Meta Language), a binary derivative of XML. EBML enables significant advantages in terms of future format extensibility, without breaking file support in old parsers.¶
First, it is essential to clarify exactly "What an Audio/Video container is", to avoid any misunderstandings:¶
It is NOT a video or audio compression format (codec)¶
It is an envelope for which there can be many audio, video and subtitles streams, allowing the user to store a complete movie or CD in a single file.¶
Matroska is designed with the future in mind. It incorporates features like:¶
Matroska is an open standards project. This means for personal use it is absolutely free to use and that the technical specifications describing the bitstream are open to everybody, even to companies that would like to support it in their products.¶
This document is a work-in-progress specification defining the Matroska file format as part of the IETF Cellar working group. But since it's quite complete it is used as a reference for the development of libmatroska. Legacy versions of the specification can be found here (PDF doc by Alexander Noe -- outdated).¶
For a simplified diagram of the layout of a Matroska file, see the Diagram page.¶
A more refined and detailed version of the EBML specifications is being worked on here.¶
The table found below is now generated from the "source" of the Matroska specification. This XML file is also used to generate the semantic data used in libmatroska and libmatroska2. We encourage anyone to use and monitor its changes so your code is spec-proof and always up to date.¶
Note that versions 1, 2 and 3 have been finalized. Version 4 is currently work in progress. There MAY be further additions to v4.¶
Storage of a arbitrary and potentially executable data within an Attachment Element. Matroska Readers that extract or use data from Matroska Attachments SHOULD check that the data adheres to expectations.¶
A Matroska Attachment with an inaccurate mime-type.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119][RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This document defines specific terms in order to define the format and application of Matroska. Specific terms are defined below:¶
Matroska: a multimedia container format based on EBML (Extensible Binary Meta Language)¶
Matroska Reader: A Matroska Reader is a data parser that interprets the semantics of a Matroska document and creates a way for programs to use Matroska.¶
Matroska Player: A Matroska Player is a Matroska Reader with a primary purpose of playing audiovisual files, including Matroska documents.¶
Matroska is a Document Type of EBML (Extensible Binary Meta Language). This specification is dependent on the EBML Specification. For an understanding of Matroska's EBML Schema, see in particular the sections of the EBML Specification covering EBML Element Types, EBML Schema, and EBML Structure.¶
Matroska from version 1 through 3 uses language codes that can be either the 3 letters bibliographic ISO-639-2 form (like "fre" for french), or such a language code followed by a dash and a country code for specialities in languages (like "fre-ca" for Canadian French). The ISO 639-2 Language Elements are "Language Element", "TagLanguage Element", and "ChapLanguage Element".¶
Starting in Matroska version 4, either ISO 639-2 or BCP 47 MAY be used, although BCP 47 is RECOMMENDED. The BCP 47 Language Elements are "LanguageIETF Element", "TagLanguageIETF Element", and "ChapLanguageIETF Element". If a BCP 47 Language Element and an ISO 639-2 Language Element are used within the same Parent Element, then the ISO 639-2 Language Element MUST be ignored and precedence given to the BCP 47 Language Element.¶
Each level can have different meanings for audio and video. The ORIGINAL_MEDIUM tag can be used to specify a string for ChapterPhysicalEquiv = 60. Here is the list of possible levels for both audio and video :¶
Frames using references SHOULD be stored in "coding order". That means the references first and then the frames referencing them. A consequence is that timestamps might not be consecutive. But a frame with a past timestamp MUST reference a frame already known, otherwise it's considered bad/void.¶
There can be many Blocks in a BlockGroup provided they all have the same timestamp. It is used with different parts of a frame with different priorities.¶
Track Number (Track Entry). It is coded in EBML like form (1 octet if the value is < 0x80, 2 if < 0x4000, etc) (most significant bits set to increase the range).
0x01+
MUST
Timestamp (relative to Cluster timestamp, signed int16)
Lacing is a mechanism to save space when storing data. It is typically used for small blocks of data (referred to as frames in Matroska). There are 3 types of lacing:¶
Xiph, inspired by what is found in the Ogg container¶
EBML, which is the same with sizes coded differently¶
For example, a user wants to store 3 frames of the same track. The first frame is 800 octets long, the second is 500 octets long and the third is 1000 octets long. As these data are small, they can be stored in a lace to save space. They will then be stored in the same block as follows:¶
Lacing head: Number of frames in the lace -1, i.e. 2 (the 800 and 500 octets one)¶
Lacing sizes: only the 2 first ones will be coded, 800 gives 255;255;255;35, 500 gives 255;245. The size of the last frame is deduced from the total size of the Block.¶
In this case, the size is not coded as blocks of 255 bytes, but as a difference with the previous size and this size is coded as in EBML. The first size in the lace is unsigned as in EBML. The others use a range shifting to get a sign on each value:¶
Lacing head: Number of frames in the lace -1, i.e. 2 (the 800 and 500 octets one)¶
Lacing sizes: only the 2 first ones will be coded, 800 gives 0x320 0x4000 = 0x4320, 500 is coded as -300 : - 0x12C + 0x1FFF + 0x4000 = 0x5ED3. The size of the last frame is deduced from the total size of the Block.¶
In this case, only the number of frames in the lace is saved, the size of each frame is deduced from the total size of the Block. For example, for 3 frames of 800 octets each:¶
The SimpleBlock is inspired by the Block structure. The main differences are the added Keyframe flag and Discardable flag. Otherwise everything is the same.¶
Frames using references SHOULD be stored in "coding order". That means the references first and then the frames referencing them. A consequence is that timestamps might not be consecutive. But a frame with a past timestamp MUST reference a frame already known, otherwise it's considered bad/void.¶
There can be many Block Elements in a BlockGroup provided they all have the same timestamp. It is used with different parts of a frame with different priorities.¶
Track Number (Track Entry). It is coded in EBML like form (1 octet if the value is < 0x80, 2 if < 0x4000, etc) (most significant bits set to increase the range).
0x01+
MUST
Timestamp (relative to Cluster timestamp, signed int16)
Lace-coded size of each frame of the lace, except for the last one (multiple uint8). *This is not used with Fixed-size lacing as it is calculated automatically from (total size of lace) / (number of frames in lace).
A Matroska file MUST be composed of at least one EBML Document using the Matroska Document Type. Each EBML Document MUST start with an EBML Header and MUST be followed by the EBML Root Element, defined as Segment in Matroska. Matroska defines several Top Level Elements which MAY occur within the Segment.¶
As an example, a simple Matroska file consisting of a single EBML Document could be represented like this:¶
The following diagram represents a simple Matroska file, comprised of an EBML Document with an EBML Header, a Segment Element (the Root Element), and all eight Matroska Top Level Elements. In the following diagrams of this section, horizontal spacing expresses a parent-child relationship between Matroska Elements (e.g. the Info Element is contained within the Segment Element) whereas vertical alignment represents the storage order within the file.¶
The Matroska EBML Schema defines eight Top Level Elements: SeekHead, Info, Tracks, Chapters, Cluster, Cues, Attachments, and Tags.¶
The SeekHead Element (also known as MetaSeek) contains an index of Top Level Elements locations within the Segment. Use of the SeekHead Element is RECOMMENDED. Without a SeekHead Element, a Matroska parser would have to search the entire file to find all of the other Top Level Elements. This is due to Matroska's flexible ordering requirements; for instance, it is acceptable for the Chapters Element to be stored after the Cluster Elements.¶
The Info Element contains vital information for identifying the whole Segment. This includes the title for the Segment, a randomly generated unique identifier, and the unique identifier(s) of any linked Segment Elements.¶
The Tracks Element defines the technical details for each track and can store the name, number, unique identifier, language and type (audio, video, subtitles, etc.) of each track. For example, the Tracks Element MAY store information about the resolution of a video track or sample rate of an audio track.¶
The Tracks Element MUST identify all the data needed by the codec to decode the data of the specified track. However, the data required is contingent on the codec used for the track. For example, a Track Element for uncompressed audio only requires the audio bit rate to be present. A codec such as AC-3 would require that the CodecID Element be present for all tracks, as it is the primary way to identify which codec to use to decode the track.¶
The Chapters Element lists all of the chapters. Chapters are a way to set predefined points to jump to in video or audio.¶
Cluster Elements contain the content for each track, e.g. video frames. A Matroska file SHOULD contain at least one Cluster Element. The Cluster Element helps to break up SimpleBlock or BlockGroup Elements and helps with seeking and error protection. It is RECOMMENDED that the size of each individual Cluster Element be limited to store no more than 5 seconds or 5 megabytes. Every Cluster Element MUST contain a Timestamp Element. This SHOULD be the Timestamp Element used to play the first Block in the Cluster Element. There SHOULD be one or more BlockGroup or SimpleBlock Element in each Cluster Element. A BlockGroup Element MAY contain a Block of data and any information relating directly to that Block.¶
Each Cluster MUST contain exactly one Timestamp Element. The Timestamp Element value MUST be stored once per Cluster. The Timestamp Element in the Cluster is relative to the entire Segment. The Timestamp Element SHOULD be the first Element in the Cluster.¶
Additionally, the Block contains an offset that, when added to the Cluster's Timestamp Element value, yields the Block's effective timestamp. Therefore, timestamp in the Block itself is relative to the Timestamp Element in the Cluster. For example, if the Timestamp Element in the Cluster is set to 10 seconds and a Block in that Cluster is supposed to be played 12 seconds into the clip, the timestamp in the Block would be set to 2 seconds.¶
The ReferenceBlock in the BlockGroup is used instead of the basic "P-frame"/"B-frame" description. Instead of simply saying that this Block depends on the Block directly before, or directly afterwards, the Timestamp of the necessary Block is used. Because there can be as many ReferenceBlock Elements as necessary for a Block, it allows for some extremely complex referencing.¶
The Cues Element is used to seek when playing back a file by providing a temporal index for some of the Tracks. It is similar to the SeekHead Element, but used for seeking to a specific time when playing back the file. It is possible to seek without this element, but it is much more difficult because a Matroska Reader would have to 'hunt and peck' through the file looking for the correct timestamp.¶
The Cues Element SHOULD contain at least one CuePoint Element. Each CuePoint Element stores the position of the Cluster that contains the BlockGroup or SimpleBlock Element. The timestamp is stored in the CueTime Element and location is stored in the CueTrackPositions Element.¶
The Cues Element is flexible. For instance, Cues Element can be used to index every single timestamp of every Block or they can be indexed selectively. For video files, it is RECOMMENDED to index at least the keyframes of the video track.¶
The Attachments Element is for attaching files to a Matroska file such as pictures, webpages, programs, or even the codec needed to play back the file.¶
The Tags Element contains metadata that describes the Segment and potentially its Tracks, Chapters, and Attachments. Each Track or Chapter that those tags applies to has its UID listed in the Tags. The Tags contain all extra information about the file: scriptwriter, singer, actors, directors, titles, edition, price, dates, genre, comments, etc. Tags can contain their values in multiple languages. For example, a movie's "title" Tag might contain both the original English title as well as the title it was released as in Germany.¶
A boolean to express if the Matroska Element is also supported within version 2 of the webm specification. Please consider the webm specification as the authoritative on webm.
This specification includes an EBML Schema which defines the Elements and structure of Matroska as an EBML Document Type. The EBML Schema defines every valid Matroska element in a manner defined by the EBML specification.¶
Here the definition of each Matroska Element is provided.¶
definition: A unique ID to identify the previous Segment of a Linked Segment (128 bits).¶
usage notes: If the Segment is a part of a Linked Segment that uses Hard Linking then either the PrevUID or the NextUID Element is REQUIRED. If a Segment contains a PrevUID but not a NextUID then it MAY be considered as the last Segment of the Linked Segment. The PrevUID MUST NOT be equal to the SegmentUID.¶
definition: A filename corresponding to the file of the previous Linked Segment.¶
usage notes: Provision of the previous filename is for display convenience, but PrevUID SHOULD be considered authoritative for identifying the previous Segment in a Linked Segment.¶
definition: A unique ID to identify the next Segment of a Linked Segment (128 bits).¶
usage notes: If the Segment is a part of a Linked Segment that uses Hard Linking then either the PrevUID or the NextUID Element is REQUIRED. If a Segment contains a NextUID but not a PrevUID then it MAY be considered as the first Segment of the Linked Segment. The NextUID MUST NOT be equal to the SegmentUID.¶
definition: A filename corresponding to the file of the next Linked Segment.¶
usage notes: Provision of the next filename is for display convenience, but NextUID SHOULD be considered authoritative for identifying the Next Segment.¶
definition: The list of tracks that are not used in that part of the stream. It is useful when using overlay tracks on seeking or to decide what track to use.¶
definition: One of the track number that are not used from now on in the stream. It could change later if not specified as silent in a further Cluster.¶
definition: Similar to Block but without all the extra information, mostly used to reduced overhead when no extra feature is needed. (see SimpleBlock Structure)¶
definition: Contain additional blocks to complete the main one. An EBML parser that has no knowledge of the Block structure could still see and use/skip these data.¶
definition: An ID to identify the BlockAdditional level. A value of 1 means the BlockAdditional data is interpreted as additional data passed to the codec with the Block data.¶
definition: The duration of the Block (based on TimestampScale). The BlockDuration Element can be useful at the end of a Track to define the duration of the last frame (as there is no subsequent Block available), or when there is a break in a track like for subtitle tracks.¶
BlockDuration MUST be set (minOccurs=1) if the associated TrackEntry stores a DefaultDuration value.
default
When not written and with no DefaultDuration, the value is assumed to be the difference between the timestamp of this Block and the timestamp of the next Block in "display" order (not coding order).
definition: This frame is referenced and has the specified cache priority. In cache only a frame of the same or higher priority can replace this frame. A value of 0 means the frame is not referenced.¶
definition: Duration in nanoseconds of the silent data added to the Block (padding at the end of the Block for positive value, at the beginning of the Block for negative value). The duration of DiscardPadding is not calculated in the duration of the TrackEntry and SHOULD be discarded during playback.¶
definition: The reverse number of the frame in the lace (0 is the last frame, 1 is the next to last, etc). Being able to interpret this Element is not REQUIRED for playback.¶
definition: The track number as used in the Block Header (using more than 127 tracks is not encouraged, though the design allows an unlimited number).¶
definition: Set if that track MUST be active during playback. There can be many forced track for a kind (audio, video or subs), the player SHOULD select the one which language matches the user preference or the default + forced track. Overlay MAY happen between a forced and non-forced track of the same kind. (1 bit)¶
definition: The minimum number of frames a player SHOULD be able to cache during playback. If set to 0, the reference pseudo-cache system is not used.¶
definition: The period in nanoseconds (not scaled by TimestampScale) between two successive fields at the output of the decoding process (see the notes)¶
definition: DEPRECATED, DO NOT USE. The scale to apply on this track to work at normal speed in relation with other tracks (mostly used to adjust video speed when the audio length differs).¶
definition: Extra binary data that the BlockAddIDType can use to interpret the BlockAdditional data. The intepretation of the binary data depends on the BlockAddIDType value and the corresponding Block Additional Mapping.¶
definition: Specifies the language of the track in the Matroska languages form. This Element MUST be ignored if the LanguageIETF Element is used in the same TrackEntry.¶
definition: Specifies the language of the track according to BCP 47 and using the IANA Language Subtag Registry. If this Element is used, then any Language Elements used in the same TrackEntry MUST be ignored.¶
definition: Specify that this track is an overlay track for the Track specified (in the u-integer). That means when this track has a gap (see SilentTracks) the overlay track SHOULD be used instead. The order of multiple TrackOverlay matters, the first one is the one that SHOULD be used. If not found it SHOULD be the second, etc.¶
definition: CodecDelay is The codec-built-in delay in nanoseconds. This value MUST be subtracted from each block timestamp in order to get the actual timestamp. The value SHOULD be small so the muxing of tracks with the same actual timestamp are in the same Cluster.¶
If the DisplayUnit of the same TrackEntry is 0, then the default value for DisplayWidth is equal to PixelWidth - PixelCropLeft - PixelCropRight, else there is no default value.
If the DisplayUnit of the same TrackEntry is 0, then the default value for DisplayHeight is equal to PixelHeight - PixelCropTop - PixelCropBottom, else there is no default value.
definition: Specify the pixel format used for the Track's data as a FourCC. This value is similar in scope to the biCompression value of AVI's BITMAPINFOHEADER.¶
definition: The Matrix Coefficients of the video used to derive luma and chroma values from red, green, and blue color primaries. For clarity, the value and meanings for MatrixCoefficients are adopted from Table 4 of ISO/IEC 23001-8:2016 or ITU-T H.273.¶
definition: The amount of pixels to remove in the Cr and Cb channels for every pixel not removed horizontally. Example: For video with 4:2:0 chroma subsampling, the ChromaSubsamplingHorz SHOULD be set to 1.¶
definition: The amount of pixels to remove in the Cr and Cb channels for every pixel not removed vertically. Example: For video with 4:2:0 chroma subsampling, the ChromaSubsamplingVert SHOULD be set to 1.¶
definition: The amount of pixels to remove in the Cb channel for every pixel not removed horizontally. This is additive with ChromaSubsamplingHorz. Example: For video with 4:2:1 chroma subsampling, the ChromaSubsamplingHorz SHOULD be set to 1 and CbSubsamplingHorz SHOULD be set to 1.¶
definition: The transfer characteristics of the video. For clarity, the value and meanings for TransferCharacteristics are adopted from Table 3 of ISO/IEC 23091-4 or ITU-T H.273.¶
definition: The colour primaries of the video. For clarity, the value and meanings for Primaries are adopted from Table 2 of ISO/IEC 23091-4 or ITU-T H.273.¶
definition: Private data that only applies to a specific projection.SemanticsIf ProjectionType equals 0 (Rectangular), then this element must not be present.If ProjectionType equals 1 (Equirectangular), then this element must be present and contain the same binary data that would be stored inside an ISOBMFF Equirectangular Projection Box ('equi').If ProjectionType equals 2 (Cubemap), then this element must be present and contain the same binary data that would be stored inside an ISOBMFF Cubemap Projection Box ('cbmp').If ProjectionType equals 3 (Mesh), then this element must be present and contain the same binary data that would be stored inside an ISOBMFF Mesh Projection Box ('mshp').Note: ISOBMFF box size and fourcc fields are not included in the binary data, but the FullBox version and flag fields are. This is to avoid redundant framing information while preserving versioning and semantics between the two container formats.¶
definition: Specifies a yaw rotation to the projection.SemanticsValue represents a clockwise rotation, in degrees, around the up vector. This rotation must be applied before any ProjectionPosePitch or ProjectionPoseRoll rotations. The value of this field should be in the -180 to 180 degree range.¶
definition: Specifies a pitch rotation to the projection.SemanticsValue represents a counter-clockwise rotation, in degrees, around the right vector. This rotation must be applied after the ProjectionPoseYaw rotation and before the ProjectionPoseRoll rotation. The value of this field should be in the -90 to 90 degree range.¶
definition: Specifies a roll rotation to the projection.SemanticsValue represents a counter-clockwise rotation, in degrees, around the forward vector. This rotation must be applied after the ProjectionPoseYaw and ProjectionPosePitch rotations. The value of this field should be in the -180 to 180 degree range.¶
definition: Operation that needs to be applied on tracks to create this virtual track. For more details look at the Specification Notes on the subject.¶
definition: Tells when this modification was used during encoding/muxing starting with 0 and counting upwards. The decoder/demuxer has to start with the highest order number it finds and work its way down. This value has to be unique over all ContentEncodingOrder Elements in the TrackEntry that contains this ContentEncodingOrder element.¶
definition: Settings describing the compression used. This Element MUST be present if the value of ContentEncodingType is 0 and absent otherwise. Each block MUST be decompressable even if no previous block is available in order not to prevent seeking.¶
definition: Settings that might be needed by the decompressor. For Header Stripping (ContentCompAlgo=3), the bytes that were removed from the beginning of each frames of the track.¶
definition: Settings describing the encryption used. This Element MUST be present if the value of ContentEncodingType is 1 (encryption) and MUST be ignored otherwise.¶
definition: The relative position inside the Cluster of the referenced SimpleBlock or BlockGroup with 0 being the first possible position for an Element inside that Cluster.¶
definition: The duration of the block according to the Segment time base. If missing the track's DefaultDuration does not apply and no duration information is available in terms of the cues.¶
definition: The Segment Position of the Codec State corresponding to this referenced Element. 0 means that the data is taken from the initial Track Entry.¶
definition: Specify whether the chapter is enabled. It can be enabled/disabled by a Control Track. When disabled, the movie SHOULD skip all the content between the TimeStart and TimeEnd of this chapter (see flag notes). (1 bit)¶
definition: The EditionUID to play from the Segment linked in ChapterSegmentUID. If ChapterSegmentEditionUID is undeclared then no Edition of the linked Segment is used.¶
definition: UID of the Track to apply this chapter too. In the absence of a control track, choosing this chapter will select the listed Tracks and deselect unlisted tracks. Absence of this Element indicates that the Chapter SHOULD be applied to any currently used Tracks.¶
definition: The languages corresponding to the string, in the bibliographic ISO-639-2 form. This Element MUST be ignored if the ChapLanguageIETF Element is used within the same ChapterDisplay Element.¶
definition: Specifies the language used in the ChapString according to BCP 47 and using the IANA Language Subtag Registry. If this Element is used, then any ChapLanguage Elements used in the same ChapterDisplay MUST be ignored.¶
definition: The countries corresponding to the string, same 2 octets as in Internet domains. This Element MUST be ignored if the ChapLanguageIETF Element is used within the same ChapterDisplay Element.¶
definition: Contains the type of the codec used for the processing. A value of 0 means native Matroska processing (to be defined), a value of 1 means the DVD command set is used. More codec IDs can be added later.¶
definition: Contains the command information. The data SHOULD be interpreted depending on the ChapProcessCodecID value. For ChapProcessCodecID = 1, the data correspond to the binary DVD cell pre/post commands.¶
definition: Element containing metadata describing Tracks, Editions, Chapters, Attachments, or the Segment as a whole. A list of valid tags can be found here.¶
definition: Specifies which other elements the metadata represented by the Tag applies to. If empty or not present, then the Tag describes everything in the Segment.¶
definition: An informational string that can be used to display the logical level of the target like "ALBUM", "TRACK", "MOVIE", "CHAPTER", etc (see TargetType).¶
definition: A unique ID to identify the EditionEntry(s) the tags belong to. If the value is 0 at this level, the tags apply to all editions in the Segment.¶
definition: A unique ID to identify the Chapter(s) the tags belong to. If the value is 0 at this level, the tags apply to all chapters in the Segment.¶
definition: A unique ID to identify the Attachment(s) the tags belong to. If the value is 0 at this level, the tags apply to all the attachments in the Segment.¶
definition: Specifies the language of the tag specified, in the Matroska languages form. This Element MUST be ignored if the TagLanguageIETF Element is used within the same SimpleTag Element.¶
definition: Specifies the language used in the TagString according to BCP 47 and using the IANA Language Subtag Registry. If this Element is used, then any TagLanguage Elements used in the same SimpleTag MUST be ignored.¶
Except for the EBML Header and the CRC-32 Element, the EBML specification does not require any particular storage order for Elements. The Matroska specification however defines mandates and recommendations for ordering certain Elements in order to facilitate better playback, seeking, and editing efficiency. This section describes and offers rationale for ordering requirements and recommendations for Matroska.¶
The Info Element is the only REQUIRED Top-Level Element in a Matroska file. To be playable, Matroska MUST also contain at least one Tracks Element and Cluster Element. The first Info Element and the first Tracks Element MUST either be stored before the first Cluster Element or both SHALL be referenced by a SeekHead Element occurring before the first Cluster Element.¶
It is possible to edit a Matroska file after it has been created. For example, chapters, tags or attachments can be added. When new Top-Level Elements are added to a Matroska file, the SeekHead Element(s) MUST be updated so that the SeekHead Element(s) itemize the identity and position of all Top-Level Elements. Editing, removing, or adding Elements to a Matroska file often requires that some existing Elements be voided or extended; therefore, it is RECOMMENDED to use Void Elements as padding in between Top-Level Elements.¶
As noted by the EBML specification, if a CRC-32 Element is used then the CRC-32 Element MUST be the first ordered Element within its Parent Element. The Matroska specification recommends that CRC-32 Elements SHOULD NOT be used as an immediate Child Element of the Segment Element; however all Top-Level Elements of an EBML Document SHOULD include a CRC-32 Element as a Child Element.¶
If used, the first SeekHead Element SHOULD be the first non-CRC-32 Child Element of the Segment Element. If a second SeekHead Element is used, then the first SeekHead Element MUST reference the identity and position of the second SeekHead. Additionally, the second SeekHead Element MUST only reference Cluster Elements and not any other Top-Level Element already contained within the first SeekHead Element. The second SeekHead Element MAY be stored in any order relative to the other Top-Level Elements. Whether one or two SeekHead Element(s) are used, the SeekHead Element(s) MUST collectively reference the identity and position of all Top-Level Elements except for the first SeekHead Element.¶
It is RECOMMENDED that the first SeekHead Element be followed by a Void Element to allow for the SeekHead Element to be expanded to cover new Top-Level Elements that could be added to the Matroska file, such as Tags, Chapters and Attachments Elements.¶
The Cues Element is RECOMMENDED to optimize seeking access in Matroska. It is programmatically simpler to add the Cues Element after all Cluster Elements have been written because this does not require a prediction of how much space to reserve before writing the Cluster Elements. However, storing the Cues Element before the Cluster Elements can provide some seeking advantages. If the Cues Element is present, then it SHOULD either be stored before the first Cluster Element or be referenced by a SeekHead Element.¶
The Chapters Element SHOULD be placed before the Cluster Element(s). The Chapters Element can be used during playback even if the user does not need to seek. It immediately gives the user information about what section is being read and what other sections are available. In the case of Ordered Chapters it RECOMMENDED to evaluate the logical linking even before playing. The Chapters Element SHOULD be placed before the first Tracks Element and after the first Info Element.¶
The Attachments Element is not intended to be used by default when playing the file, but could contain information relevant to the content, such as cover art or fonts. Cover art is useful even before the file is played and fonts could be needed before playback starts for initialization of subtitles. The Attachments Element MAY be placed before the first Cluster Element; however if the Attachments Element is likely to be edited, then it SHOULD be placed after the last Cluster Element.¶
The Tags Element is most subject to changes after the file was originally created. For easier editing, the Tags Element SHOULD be placed at the end of the Segment Element, even after the Attachments Element. On the other hand, it is inconvenient to have to seek in the Segment for tags, especially for network streams. So it's better if the Tags Element is found early in the stream. When editing the Tags Element, the original Tags Element at the beginning can be overwritten with a Void Element and a new Tags Element written at the end of the Segment Element. The file size will only marginally change.¶
Two Chapter Flags are defined to describe the behavior of the ChapterAtom Element: ChapterFlagHidden and ChapterFlagEnabled.¶
If a ChapterAtom Element is the Child Element of another ChapterAtom Element with a Chapter Flag set to true, then the Child ChapterAtom Element MUST be interpreted as having its same Chapter Flag set to true. If a ChapterAtom Element is the Child Element of another ChapterAtom Element with a Chapter Flag set to false or if the ChapterAtom Element does not have a ChapterAtom Element as its Parent Element, then it MUST be interpreted according to its own Chapter Flag.¶
As an example, consider a Parent ChapterAtom Element that has its ChapterFlagHidden set to true and also contains two child ChapterAtoms, the first with ChapterFlagHidden set to true and the second with ChapterFlagHidden either set to false or not present at all (in which case the default value of the Element applies, which is false). Since the parent ChapterAtom has its ChapterFlagHidden set to true then all of its children ChapterAtoms MUST also be interpreted as if their ChapterFlagHidden is also set to true. However, if a Control Track toggles the parent's ChapterFlagHidden flag to false, then only the parent ChapterAtom and its second child ChapterAtom MUST be interpreted as if ChapterFlagHidden is set to false. The first child ChapterAtom which has the ChapterFlagHidden flag set to true retains its value until its value is toggled to false by a Control Track.¶
The EditionFlagHidden Flag behaves similarly to the ChapterFlagHidden Flag: if EditionFlagHidden is set to true, its Child ChapterAtoms Elements MUST also be interpreted as if their ChapterFlagHidden is also set to true, regardless of their own ChapterFlagHidden Flags. If EditionFlagHidden is toggled by a Control Track to false, the ChapterFlagHidden Flags of the Child ChapterAtoms Elements SHALL determine whether the ChapterAtom is hidden or not.¶
It is RECOMMENDED that no more than one Edition have an EditionFlagDefault Flag set to true. The first Edition with both the EditionFlagDefault Flag set to true and the EditionFlagHidden Flag set to false is the Default Edition. When all EditionFlagDefault Flags are set to false, then the first Edition is the Default Edition.¶
The EditionFlagOrdered Flag is a significant feature as it enables an Edition of Ordered Chapters which defines and arranges a virtual timeline rather than simply labeling points within the timeline. For example, with Editions of Ordered Chapters a single Matroska file can present multiple edits of a film without duplicating content. Alternatively if a videotape is digitized in full, one Ordered Edition could present the full content (including colorbars, countdown, slate, a feature presentation, and black frames), while another Edition of Ordered Chapters can use Chapters that only mark the intended presentation with the colorbars and other ancillary visual information excluded. If an Edition of Ordered Chapters is enabled then the Matroska Player MUST play those Chapters in their stored order from the timestamp marked in the ChapterTimeStart Element to the timestamp marked in to ChapterTimeEnd Element.¶
If the EditionFlagOrdered Flag is set to false, Simple Chapters are used and only the ChapterTimeStart of a Chapter is used as chapter mark to jump to the predefined point in the timeline. With Simple Chapters, a Matroska Player MUST ignore certain Chapter Elements. All these elements are now informational only.¶
The following list shows the different usage of Chapter Elements between an ordered and non-ordered Edition.¶
Chapter elements / ordered Edition | False | True ChapterUID | X | X ChapterStringUID | X | X ChapterTimeStart | X | X ChapterTimeEnd | - | X ChapterFlagHidden | X | X ChapterFlagEnabled | X | X ChapterSegmentUID | - | X ChapterSegmentEditionUID | - | X ChapterPhysicalEquiv | X | X ChapterTrack | - | X ChapterDisplay | X | X ChapProcess | - | X¶
Furthermore there are other EBML Elements which could be used if the EditionFlagOrdered Flag is set to true.¶
Other elements / ordered Edition | False | True Info/SegmentFamily | - | X Info/ChapterTranslate | - | X Track/TrackTranslate | - | X¶
These other Elements belong to the Matroska DVD menu system and are only used when the ChapProcessCodecID Element is set to 1.¶
Hard Linking: Ordered-Chapters supersedes the Hard Linking.¶
Soft Linking: In this complex system Ordered Chapters are REQUIRED and a Chapter CODEC MUST interpret the ChapProcess of all chapters.¶
Medium Linking: Ordered Chapters are used in a normal way and can be combined with the ChapterSegmentUID element which establishes a link to another Matroska file/Segment.¶
The menu features are handled like a chapter codec. That means each codec has a type, some private data and some data in the chapters.¶
The type of the menu system is defined by the ChapProcessCodecID parameter. For now only 2 values are supported : 0 matroska script, 1 menu borrowed from the DVD. The private data depend on the type of menu system (stored in ChapProcessPrivate), idem for the data in the chapters (stored in ChapProcessData).¶
This is the case when ChapProcessCodecID = 0. This is a script language build for Matroska purposes. The inspiration comes from ActionScript, javascript and other similar scripting languages. The commands are stored as text commands, in UTF-8. The syntax is C like, with commands spanned on many lines, each terminating with a ";". You can also include comments at the end of lines with "//" or comment many lines using "/* */". The scripts are stored in ChapProcessData. For the moment ChapProcessPrivate is not used.¶
The one and only command existing for the moment is GotoAndPlay( ChapterUID );. As the same suggests, it means that when this command is encountered, the Matroska Player SHOULD jump to the Chapter specified by the UID and play it.¶
This is the case when ChapProcessCodecID = 1. Each level of a chapter corresponds to a logical level in the DVD system that is stored in the first octet of the ChapProcessPrivate. This DVD hierarchy is as follows:¶
ChapProcessPrivate | DVD Name | Hierarchy | Commands Possible | Comment 0x30 | SS | DVD domain | - | First Play, Video Manager, Video Title 0x2A | LU | Language Unit | - | Contains only PGCs 0x28 | TT | Title | - | Contains only PGCs 0x20 | PGC | Program Group Chain (PGC) | * | 0x18 | PG | Program 1 / Program 2 / Program 3 | - | 0x10 | PTT | Part Of Title 1 / Part Of Title 2 | - | Equivalent to the chapters on the sleeve. 0x08 | CN | Cell 1 / Cell 2 / Cell 3 / Cell 4 / Cell 5 / Cell 6 | - |¶
You can also recover wether a Segment is a Video Manager (VMG), Video Title Set (VTS) or Video Title Set Menu (VTSM) from the ChapterTranslateID element found in the Segment Info. This field uses 2 octets as follows:¶
Domain Type: 0 for VMG, the domain number for VTS and VTSM¶
Domain Value: 0 for VMG and VTSM, 1 for the VTS source.¶
For instance, the menu part from VTS010.VOB would be coded [1,0] and the content part from VTS023.VOB would be [2,1]. The VMG is always [0,0]¶
The following octets of ChapProcessPrivate are as follows:¶
Octet 1 | DVD Name | Following Octets 0x30 | SS | Domain name code (1: 0x00= First play, 0xC0= VMG, 0x40= VTSM, 0x80= VTS) + VTS(M) number (2) 0x2A | LU | Language code (2) + Language extension (1) 0x28 | TT | global Title number (2) + corresponding TTN of the VTS (1) 0x20 | PGC | PGC number (2) + Playback Type (1) + Disabled User Operations (4) 0x18 | PG | Program number (2) 0x10 | PTT | PTT-chapter number (1) 0x08 | CN | Cell number [VOB ID(2)][Cell ID(1)][Angle Num(1)]¶
If the level specified in ChapProcessPrivate is a PGC (0x20), there is an octet called the Playback Type, specifying the kind of PGC defined:¶
0x87: Chapter Menu (only found in the VTSM domain)¶
The next 4 following octets correspond to the User Operation flags in the standard PGC. When a bit is set, the command SHOULD be disabled.¶
ChapProcessData contains the pre/post/cell commands in binary format as there are stored on a DVD. There is just an octet preceding these data to specify the number of commands in the element. As follows: [# of commands(1)][command 1 (8)][command 2 (8)][command 3 (8)].¶
More information on the DVD commands and format on DVD-replica, where we got most of the info about it. You can also get information on DVD from the DVDinfo project.¶
Matroska supports storage of related files and data in the Attachments Element (a Top-Level Element). Attachment Elements can be used to store related cover art, font files, transcripts, reports, error recovery files, picture or text-based annotations, copies of specifications, or other ancillary files related to the Segment.¶
Matroska Readers MUST NOT execute files stored as Attachment Elements.¶
This section defines a set of guidelines for the storage of cover art in Matroska files. A Matroska Reader MAY use embedded cover art to display a representational still-image depiction of the multimedia contents of the Matroska file.¶
Only JPEG and PNG image formats SHOULD be used for cover art pictures.¶
There can be two different covers for a movie/album: a portrait style (e.g., a DVD case) and a landscape style (e.g., a wide banner ad).¶
There can be two versions of the same cover, the normal cover and the small cover. The dimension of the normal cover SHOULD be 600 pixels on the smallest side (for example, 960x600 for landscape, 600x800 for portrait, or 600x600 for square). The dimension of the small cover SHOULD be 120 pixels on the smallest side (for example, 192x120 or 120x160).¶
Versions of cover art can be differentiated by the filename, which is stored in the FileName Element. The default filename of the normal cover in square or portrait mode is cover.(jpg|png). When stored, the normal cover SHOULD be the first Attachment in storage order. The small cover SHOULD be prefixed with "small_", such as small_cover.(jpg|png). The landscape variant SHOULD be suffixed with "_land", such as cover_land.(jpg|png). The filenames are case sensitive.¶
The following table provides examples of file names for cover art in Attachments.¶
FileName | Image Orientation | Pixel Length of Smallest Side cover.jpg | Portrait or square | 600 smallcover.png | Portrait or square | 120 coverland.png | Landscape | 600 smallcoverland.jpg | Landscape | 120¶
The Cues Element provides an index of certain Cluster Elements to allow for optimized seeking to absolute timestamps within the Segment. The Cues Element contains one or many CuePoint Elements which each MUST reference an absolute timestamp (via the CueTime Element), a Track (via the CueTrack Element), and a Segment Position (via the CueClusterPosition Element). Additional non-mandated Elements are part of the CuePoint Element such as CueDuration, CueRelativePosition, CueCodecState and others which provide any Matroska Reader with additional information to use in the optimization of seeking performance.¶
The following recommendations are provided to optimize Matroska performance.¶
Unless Matroska is used as a live stream, it SHOULD contain a Cues Element.¶
For each video track, each keyframe SHOULD be referenced by a CuePoint Element.¶
It is RECOMMENDED to not reference non-keyframes of video tracks in Cues unless it references a Cluster Element which contains a CodecState Element but no keyframes.¶
For each subtitle track present, each subtitle frame SHOULD be referenced by a CuePoint Element with a CueDuration Element.¶
References to audio tracks MAY be skipped in CuePoint Elements if a video track is present. When included the CuePoint Elements SHOULD reference audio keyframes at most once every 500 milliseconds.¶
If the referenced frame is not stored within the first SimpleBlock or first BlockGroup within its Cluster Element, then the CueRelativePosition Element SHOULD be written to reference where in the Cluster the reference frame is stored.¶
If a CuePoint Element references Cluster Element that includes a CodecState Element, then that CuePoint Element MUST use a CueCodecState Element.¶
CuePoint Elements SHOULD be numerically sorted in storage order by the value of the CueTime Element.¶
File access can simply be reading a file located on your computer, but also includes accessing a file from an HTTP (web) server or CIFS (Windows share) server. These protocols are usually safe from reading errors and seeking in the stream is possible. However, when a file is stored far away or on a slow server, seeking can be an expensive operation and SHOULD be avoided. The following guidelines, when followed, help reduce the number of seeking operations for regular playback and also have the playback start quickly without a lot of data needed to read first (like a Cues Element, Attachment Element or SeekHead Element).¶
Matroska, having a small overhead, is well suited for storing music/videos on file servers without a big impact on the bandwidth used. Matroska does not require the index to be loaded before playing, which allows playback to start very quickly. The index can be loaded only when seeking is requested the first time.¶
Livestreaming is the equivalent of television broadcasting on the internet. There are 2 families of servers for livestreaming: RTP/RTSP and HTTP. Matroska is not meant to be used over RTP. RTP already has timing and channel mechanisms that would be wasted if doubled in Matroska. Additionally, having the same information at the RTP and Matroska level would be a source of confusion if they do not match. Livestreaming of Matroska over HTTP (or any other plain protocol based on TCP) is possible.¶
A live Matroska stream is different from a file because it usually has no known end (only ending when the client disconnects). For this, all bits of the "size" portion of the Segment Element MUST be set to 1. Another option is to concatenate Segment Elements with known sizes, one after the other. This solution allows a change of codec/resolution between each segment. For example, this allows for a switch between 4:3 and 16:9 in a television program.¶
When Segment Elements are continuous, certain Elements, like MetaSeek, Cues, Chapters, and Attachments, MUST NOT be used.¶
It is possible for a Matroska Player to detect that a stream is not seekable. If the stream has neither a MetaSeek list or a Cues list at the beginning of the stream, it SHOULD be considered non-seekable. Even though it is possible to seek blindly forward in the stream, it is NOT RECOMMENDED.¶
In the context of live radio or web TV, it is possible to "tag" the content while it is playing. The Tags Element can be placed between Clusters each time it is necessary. In that case, the new Tags Element MUST reset the previously encountered Tags Elements and use the new values instead.¶
This document is a draft of the Menu system that will be the default one in Matroska. As it will just be composed of a Control Track, it will be seen as a "codec" and could be replaced later by something else if needed.¶
A menu is like what you see on DVDs, when you have some screens to select the audio format, subtitles or scene selection.¶
What we'll try to have is a system that can do almost everything done on a DVD, or more, or better, or drop the unused features if necessary.¶
As the name suggests, a Control Track is a track that can control the playback of the file and/or all the playback features. To make it as simple as possible for Matroska Players, the Control Track will just give orders to the Matroska Player and get the actions associated with the highlights/hotspots.¶
A highlight is basically a rectangle/key associated with an action UID. When that rectangle/key is activated, the Matroska Player send the UID of the action to the Control Track handler (codec). The fact that it can also be a key means that even for audio only files, a keyboard shortcut or button panel could be used for menus. But in that case, the hotspot will have to be associated with a name to display.¶
This highlight is sent from the Control Track to the Matroska Player. Then the Matroska Player has to handle that highlight until it's deactivated (see Playback Features).¶
The highlight contains a UID of the action, a displayable name (UTF-8), an associated key (list of keys to be defined, probably up/down/left/right/select), a screen position/range and an image to display. The image will be displayed either when the user place the mouse over the rectangle (or any other shape), or when an option of the screen is selected (not activated). There could be a second image used when the option is activated. And there could be a third image that can serve as background. This way you could have a still image (like in some DVDs) for the menu and behind that image blank video (small bitrate).¶
When a highlight is activated by the user, the Matroska Player has to send the UID of the action to the Control Track. Then the Control Track codec will handle the action and possibly give new orders to the Matroska Player.¶
The format used for storing images SHOULD be extensible. For the moment we'll use PNG and BMP, both with alpha channel.¶
All the actions will be written in a normal Matroska track, with a timestamp. A "Menu Frame" SHOULD be able to contain more that one action/highlight for a given timestamp. (to be determined, EBML format structure)¶
Some Matroska Players might not support the control track. That mean they will play the active/looped parts as part of the data. So I suggest putting the active/looped parts of a movie at the end of a movie. When a Menu-aware Matroska Player encounter the default Control Track of a Matroska file, the first order SHOULD be to jump at the start of the active/looped part of the movie.¶
Matroska is based upon the principle that a reading application does not have to support 100% of the specifications in order to be able to play the file. A Matroska file therefore contains version indicators that tell a reading application what to expect.¶
It is possible and valid to have the version fields indicate that the file contains Matroska Elements from a higher specification version number while signaling that a reading application MUST only support a lower version number properly in order to play it back (possibly with a reduced feature set). For example, a reading application supporting at least Matroska version V reading a file whose DocTypeReadVersion field is equal to or lower than V MUST skip Matroska/EBML Elements it encounters but does not know about if that unknown element fits into the size constraints set by the current Parent Element.¶
The default value of an Element is assumed when not present in the data stream. It is assumed only in the scope of its Parent Element. For example, the Language Element is in the scope of the Track Element. If the Parent Element is not present or assumed, then the Child Element cannot be assumed.¶
The DefaultDecodedFieldDuration Element can signal to the displaying application how often fields of a video sequence will be available for displaying. It can be used for both interlaced and progressive content. If the video sequence is signaled as interlaced, then the period between two successive fields at the output of the decoding process equals DefaultDecodedFieldDuration.¶
For video sequences signaled as progressive, it is twice the value of DefaultDecodedFieldDuration.¶
These values are valid at the end of the decoding process before post-processing (such as deinterlacing or inverse telecine) is applied.¶
Encryption in Matroska is designed in a very generic style to allow people to implement whatever form of encryption is best for them. It is possible to use the encryption framework in Matroska as a type of DRM (Digital Rights Management).¶
Because encryption occurs within the Block Element, it is possible to manipulate encrypted streams without decrypting them. The streams could potentially be copied, deleted, cut, appended, or any number of other possible editing techniques without decryption. The data can be used without having to expose it or go through the decrypting process.¶
Encryption can also be layered within Matroska. This means that two completely different types of encryption can be used, requiring two separate keys to be able to decrypt a stream.¶
Encryption information is stored in the ContentEncodings Element under the ContentEncryption Element.¶
The PixelCrop Elements (PixelCropTop, PixelCropBottom, PixelCropRight and PixelCropLeft) indicate when and by how much encoded videos frames SHOULD be cropped for display. These Elements allow edges of the frame that are not intended for display, such as the sprockets of a full-frame film scan or the VANC area of a digitized analog videotape, to be stored but hidden. PixelCropTop and PixelCropBottom store an integer of how many rows of pixels SHOULD be cropped from the top and bottom of the image (respectively). PixelCropLeft and PixelCropRight store an integer of how many columns of pixels SHOULD be cropped from the left and right of the image (respectively). For example, a pillar-boxed video that stores a 1440x1080 visual image within the center of a padded 1920x1080 encoded image MAY set both PixelCropLeft and PixelCropRight to 240, so that a Matroska Player SHOULD crop off 240 columns of pixels from the left and right of the encoded image to present the image with the pillar-boxes hidden.¶
The ProjectionPoseRoll Element (see Section 9.3.4.1.31.49) can be used to indicate that the image from the associated video track SHOULD be rotated for presentation. For instance, the following representation of the Projection Element Section 9.3.4.1.31.44) and the ProjectionPoseRoll Element represents a video track where the image SHOULD be presentation with a 90 degree counter-clockwise rotation.¶
The EBML Header of each Matroska document informs the reading application on what version of Matroska to expect. The Elements within EBML Header with jurisdiction over this information are DocTypeVersion and DocTypeReadVersion.¶
DocTypeVersion MUST be equal to or greater than the highest Matroska version number of any Element present in the Matroska file. For example, a file using the SimpleBlock Element MUST have a DocTypeVersion equal to or greater than 2. A file containing CueRelativePosition Elements MUST have a DocTypeVersion equal to or greater than 4.¶
The DocTypeReadVersion MUST contain the minimum version number that a reading application can minimally support in order to play the file back -- optionally with a reduced feature set. For example, if a file contains only Elements of version 2 or lower except for CueRelativePosition (which is a version 4 Matroska Element), then DocTypeReadVersion SHOULD still be set to 2 and not 4 because evaluating CueRelativePosition is not necessary for standard playback -- it makes seeking more precise if used.¶
DocTypeVersion MUST always be equal to or greater than DocTypeReadVersion.¶
A reading application supporting Matroska version V MUST NOT refuse to read an application with DocReadTypeVersion equal to or lower than V even if DocTypeVersion is greater than V. See also the note about Unknown Elements.¶
The Segment Position of an Element refers to the position of the first octet of the Element ID of that Element, measured in octets, from the beginning of the Element Data section of the containing Segment Element. In other words, the Segment Position of an Element is the distance in octets from the beginning of its containing Segment Element minus the size of the Element ID and Element Data Size of that Segment Element. The Segment Position of the first Child Element of the Segment Element is 0. An Element which is not stored within a Segment Element, such as the Elements of the EBML Header, do not have a Segment Position.¶
This table presents an example of Segment Position by showing a hexadecimal representation of a very small Matroska file with labels to show the offsets in octets. The file contains a Segment Element with an Element ID of 0x18538067 and a MuxingApp Element with an Element ID of 0x4D80.¶
In the above example, the Element ID of the Segment Element is stored at offset 16, the Element Data Size of the Segment Element is stored at offset 20, and the Element Data of the Segment Element is stored at offset 21.¶
The MuxingApp Element is stored at offset 26. Since the Segment Position of an Element is calculated by subtracting the position of the Element Data of the containing Segment Element from the position of that Element, the Segment Position of MuxingApp Element in the above example is 26 - 21 or 5.¶
Matroska provides several methods to link two or many Segment Elements together to create a Linked Segment. A Linked Segment is a set of multiple Segments related together into a single presentation by using Hard Linking, Medium Linking, or Soft Linking. All Segments within a Linked Segment MUST utilize the same track numbers and timescale. All Segments within a Linked Segment MUST be stored within the same directory. All Segments within a Linked Segment MUST store a SegmentUID.¶
Hard Linking (also called splitting) is the process of creating a Linked Segment by relating multiple Segment Elements using the NextUID and PrevUID Elements. Within a Linked Segment, the timestamps of each Segment MUST follow consecutively in linking order. With Hard Linking, the chapters of any Segment within the Linked Segment MUST only reference the current Segment. With Hard Linking, the NextUID and PrevUID MUST reference the respective SegmentUID values of the next and previous Segments. The first Segment of a Linked Segment SHOULD have a NextUID Element and MUST NOT have a PrevUID Element. The last Segment of a Linked Segment SHOULD have a PrevUID Element and MUST NOT have a NextUID Element. The middle Segments of a Linked Segment SHOULD have both a NextUID Element and a PrevUID Element.¶
In a chain of Linked Segments the NextUID always takes precedence over the PrevUID. So if SegmentA has a NextUID to SegmentB and SegmentB has a PrevUID to SegmentC, the link to use is SegmentA to SegmentB. If SegmentB has a PrevUID to SegmentA but SegmentA has no NextUID, then the Matroska Player MAY consider these two Segments linked as SegmentA followed by SegmentB.¶
As an example, three Segments can be Hard Linked as a Linked Segment through cross-referencing each other with SegmentUID, PrevUID, and NextUID, as in this table.¶
Medium Linking creates relationships between Segments using Ordered Chapters and the ChapterSegmentUID Element. A Segment Edition with Ordered Chapters MAY contain Chapter Elements that reference timestamp ranges from other Segments. The Segment referenced by the Ordered Chapter via the ChapterSegmentUID Element SHOULD be played as part of a Linked Segment. The timestamps of Segment content referenced by Ordered Chapters MUST be adjusted according to the cumulative duration of the the previous Ordered Chapters.¶
As an example a file named intro.mkv could have a SegmentUID of 0xb16a58609fc7e60653a60c984fc11ead. Another file called program.mkv could use a Chapter Edition that contains two Ordered Chapters. The first chapter references the Segment of intro.mkv with the use of a ChapterSegmentUID, ChapterSegmentEditionUID, ChapterTimeStart and optionally a ChapterTimeEnd element. The second chapter references content within the Segment of program.mkv. A Matroska Player SHOULD recognize the Linked Segment created by the use of ChapterSegmentUID in an enabled Edition and present the reference content of the two Segments together.¶
Soft Linking is used by codec chapters. They can reference another Segment and jump to that Segment. The way the Segments are described are internal to the chapter codec and unknown to the Matroska level. But there are Elements within the Info Element (such as ChapterTranslate) that can translate a value representing a Segment in the chapter codec and to the current SegmentUID. All Segments that could be used in a Linked Segment in this way SHOULD be marked as members of the same family via the SegmentFamily Element, so that the Matroska Player can quickly switch from one to the other.¶
The "default track" flag is a hint for a Matroska Player and SHOULD always be changeable by the user. If the user wants to see or hear a track of a certain kind (audio, video, subtitles) and hasn't chosen a specific track, the Matroska Player SHOULD use the first track of that kind whose "default track" flag is set to "1". If no such track is found then the first track of this kind SHOULD be chosen.¶
Only one track of a kind MAY have its "default track" flag set in a segment. If a track entry does not contain the "default track" flag element then its default value "1" is to be used.¶
The "forced" flag tells the Matroska Player that it MUST display/play this track or another track of the same kind that also has its "forced" flag set. When there are multiple "forced" tracks, the Matroska Player SHOULD determine the track based upon the language of the forced flag or use the default flag if no track matches the use languages. Another track of the same kind without the "forced" flag may be use simultaneously with the "forced" track (like DVD subtitles for example).¶
TrackOperation allows combining multiple tracks to make a virtual one. It uses two separate system to combine tracks. One to create a 3D "composition" (left/right/background planes) and one to simplify join two tracks together to make a single track.¶
A track created with TrackOperation is a proper track with a UID and all its flags. However the codec ID is meaningless because each "sub" track needs to be decoded by its own decoder before the "operation" is applied. The Cues Elements corresponding to such a virtual track SHOULD be the sum of the Cues Elements for each of the tracks it's composed of (when the Cues are defined per track).¶
In the case of TrackJoinBlocks, the Block Elements (from BlockGroup and SimpleBlock) of all the tracks SHOULD be used as if they were defined for this new virtual Track. When two Block Elements have overlapping start or end timestamps, it's up to the underlying system to either drop some of these frames or render them the way they overlap. This situation SHOULD be avoided when creating such tracks as you can never be sure of the end result on different platforms.¶
Overlay tracks SHOULD be rendered in the same 'channel' as the track its linked to. When content is found in such a track, it SHOULD be played on the rendering channel instead of the original track.¶
There are two different ways to compress 3D videos: have each 'eye' track in a separate track and have one track have both 'eyes' combined inside (which is more efficient, compression-wise). Matroska supports both ways.¶
For the single track variant, there is the StereoMode Element which defines how planes are assembled in the track (mono or left-right combined). Odd values of StereoMode means the left plane comes first for more convenient reading. The pixel count of the track (PixelWidth/PixelHeight) is the raw amount of pixels (for example 3840x1080 for full HD side by side) and the DisplayWidth/DisplayHeight in pixels is the amount of pixels for one plane (1920x1080 for that full HD stream). Old stereo 3D were displayed using anaglyph (cyan and red colours separated). For compatibility with such movies, there is a value of the StereoMode that corresponds to AnaGlyph.¶
There is also a "packed" mode (values 13 and 14) which consists of packing two frames together in a Block using lacing. The first frame is the left eye and the other frame is the right eye (or vice versa). The frames SHOULD be decoded in that order and are possibly dependent on each other (P and B frames).¶
For separate tracks, Matroska needs to define exactly which track does what. TrackOperation with TrackCombinePlanes do that. For more details look at how TrackOperation works.¶
The 3D support is still in infancy and may evolve to support more features.¶
The StereoMode used to be part of Matroska v2 but it didn't meet the requirement for multiple tracks. There was also a bug in libmatroska prior to 0.9.0 that would save/read it as 0x53B9 instead of 0x53B8. Matroska Readers may support these legacy files by checking Matroska v2 or 0x53B9. The older values were 0: mono, 1: right eye, 2: left eye, 3: both eyes.¶
Historically timestamps in Matroska were mistakenly called timecodes. The Timestamp Element was called Timecode, the TimestampScale Element was called TimecodeScale, the TrackTimestampScale Element was called TrackTimecodeScale and the ReferenceTimestamp Element was called ReferenceTimeCode.¶
The Block Element's timestamp MUST be a signed integer that represents the Raw Timestamp relative to the Cluster's Timestamp Element, multiplied by the TimestampScale Element. See TimestampScale for more information.¶
The Block Element's timestamp MUST be represented by a 16bit signed integer (sint16). The Block's timestamp has a range of -32768 to +32767 units. When using the default value of the TimestampScale Element, each integer represents 1ms. The maximum time span of Block Elements in a Cluster using the default TimestampScale Element of 1ms is 65536ms.¶
If a Cluster's Timestamp Element is set to zero, it is possible to have Block Elements with a negative Raw Timestamp. Block Elements with a negative Raw Timestamp are not valid.¶
The exact time of an object SHOULD be represented in nanoseconds. To find out a Block's Raw Timestamp, you need the Block's Timestamp Element, the Cluster's Timestamp Element, and the TimestampScale Element.¶
The TimestampScale Element is used to calculate the Raw Timestamp of a Block. The timestamp is obtained by adding the Block's timestamp to the Cluster's Timestamp Element, and then multiplying that result by the TimestampScale. The result will be the Block's Raw Timestamp in nanoseconds. The formula for this would look like:¶
(a + b) * c
a = `Block`'s Timestamp
b = `Cluster`'s Timestamp
c = `TimestampScale`
For example, assume a Cluster's Timestamp has a value of 564264, the Block has a Timestamp of 1233, and the TimestampScale Element is the default of 1000000.¶
Because the default value of TimestampScale is 1000000, which makes each integer in the Cluster and BlockTimestamp Elements equal 1ms, this is the most commonly used. When dealing with audio, this causes inaccuracy when seeking. When the audio is combined with video, this is not an issue. For most cases, the the synch of audio to video does not need to be more than 1ms accurate. This becomes obvious when one considers that sound will take 2-3ms to travel a single meter, so distance from your speakers will have a greater effect on audio/visual synch than this.¶
However, when dealing with audio-only files, seeking accuracy can become critical. For instance, when storing a whole CD in a single track, a user will want to be able to seek to the exact sample that a song begins at. If seeking a few sample ahead or behind, a 'crack' or 'pop' may result as a few odd samples are rendered. Also, when performing precise editing, it may be very useful to have the audio accuracy down to a single sample.¶
When storing timestamps for an audio stream, the TimestampScale Element SHOULD have an accuracy of at least that of the audio sample rate, otherwise there are rounding errors that prevent users from knowing the precise location of a sample. Here's how a program has to round each timestamp in order to be able to recreate the sample number accurately.¶
Let's assume that the application has an audio track with a sample rate of 44100. As written above the TimestampScale MUST have at least the accuracy of the sample rate itself: 1000000000 / 44100 = 22675.7369614512. This value MUST always be truncated. Otherwise the accuracy will not suffice. So in this example the application will use 22675 for the TimestampScale. The application could even use some lower value like 22674 which would allow it to be a little bit imprecise about the original timestamps. But more about that in a minute.¶
Next the application wants to write sample number 52340 and calculates the timestamp. This is easy. In order to calculate the Raw Timestamp in ns all it has to do is calculate Raw Timestamp = round(1000000000 * sample_number / sample_rate). Rounding at this stage is very important! The application might skip it if it choses a slightly smaller value for the TimestampScale factor instead of the truncated one like shown above. Otherwise it has to round or the results won't be reversible. For our example we get Raw Timestamp = round(1000000000 * 52340 / 44100) = round(1186848072.56236) = 1186848073.¶
The next step is to calculate the Absolute Timestamp - that is the timestamp that will be stored in the Matroska file. Here the application has to divide the Raw Timestamp from the previous paragraph by the TimestampScale factor and round the result: Absolute Timestamp = round(Raw Timestamp / TimestampScale_factor) which will result in the following for our example: Absolute Timestamp = round(1186848073 / 22675) = round(52341.7011245866) = 52342. This number is the one the application has to write to the file.¶
Now our file is complete, and we want to play it back with another application. Its task is to find out which sample the first application wrote into the file. So it starts reading the Matroska file and finds the TimestampScale factor 22675 and the audio sample rate 44100. Later it finds a data block with the Absolute Timestamp of 52342. But how does it get the sample number from these numbers?¶
First it has to calculate the Raw Timestamp of the block it has just read. Here's no rounding involved, just an integer multiplication: Raw Timestamp = Absolute Timestamp * TimestampScale_factor. In our example: Raw Timestamp = 52342 * 22675 = 1186854850.¶
The conversion from the Raw Timestamp to the sample number again requires rounding: sample_number = round(Raw Timestamp * sample_rate / 1000000000). In our example: sample_number = round(1186854850 * 44100 / 1000000000) = round(52340.298885) = 52340. This is exactly the sample number that the previous program started with.¶
Always calculate the timestamps / sample numbers with floating point numbers of at least 64bit precision (called 'double' in most modern programming languages). If you're calculating with integers then make sure they're 64bit long, too.¶
Always round if you divide. Always! If you don't you'll end up with situations in which you have a timestamp in the Matroska file that does not correspond to the sample number that it started with. Using a slightly lower timestamp scale factor can help here in that it removes the need for proper rounding in the conversion from sample number to Raw Timestamp.¶
The TrackTimestampScale Element is used align tracks that would otherwise be played at different speeds. An example of this would be if you have a film that was originally recorded at 24fps video. When playing this back through a PAL broadcasting system, it is standard to speed up the film to 25fps to match the 25fps display speed of the PAL broadcasting standard. However, when broadcasting the video through NTSC, it is typical to leave the film at its original speed. If you wanted to make a single file where there was one video stream, and an audio stream used from the PAL broadcast, as well as an audio stream used from the NTSC broadcast, you would have the problem that the PAL audio stream would be 1/24th faster than the NTSC audio stream, quickly leading to problems. It is possible to stretch out the PAL audio track and re-encode it at a slower speed, however when dealing with lossy audio codecs, this often results in a loss of audio quality and/or larger file sizes.¶
This is the type of problem that TrackTimestampScale was designed to fix. Using it, the video can be played back at a speed that will synch with either the NTSC or the PAL audio stream, depending on which is being used for playback. To continue the above example:¶
Track 1: Video
Track 2: NTSC Audio
Track 3: PAL Audio
Because the NTSC track is at the original speed, it will used as the default value of 1.0 for its TrackTimestampScale. The video will also be aligned to the NTSC track with the default value of 1.0.¶
The TrackTimestampScale value to use for the PAL track would be calculated by determining how much faster the PAL track is than the NTSC track. In this case, because we know the video for the NTSC audio is being played back at 24fps and the video for the PAL audio is being played back at 25fps, the calculation would be:¶
When writing a file that uses a non-default TrackTimestampScale, the values of the Block's timestamp are whatever they would be when normally storing the track with a default value for the TrackTimestampScale. However, the data is interleaved a little differently. Data SHOULD be interleaved by its Raw Timestamp in the order handed back from the encoder. The Raw Timestamp of a Block from a track using TrackTimestampScale is calculated using:¶
So, a Block from the PAL track above that had a Scaled Timestamp of 100 seconds would have a Raw Timestamp of 104.66666667 seconds, and so would be stored in that part of the file.¶
When playing back a track using the TrackTimestampScale, if the track is being played by itself, there is no need to scale it. From the above example, when playing the Video with the NTSC Audio, neither are scaled. However, when playing back the Video with the PAL Audio, the timestamps from the PAL Audio track are scaled using the TrackTimestampScale, resulting in the video playing back in synch with the audio.¶
It would be possible for a Matroska Player to also adjust the audio's samplerate at the same time as adjusting the timestamps if you wanted to play the two audio streams synchronously. It would also be possible to adjust the video to match the audio's speed. However, for playback, the selected track(s) timestamps SHOULD be adjusted if they need to be scaled.¶
While the above example deals specifically with audio tracks, this element can be used to align video, audio, subtitles, or any other type of track contained in a Matroska file.¶
Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/info/rfc2119>.
[RFC8174]
Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/info/rfc8174>.