AV1 Codec ISO Media File Format Binding

1. Bitstream features overview

OBUs are made of a 1 or 2 bytes header, identifying in particular the type of OBU, followed by an optional length field and by an optional payload, whose presence and content depend on the OBU type. Depending on its type, an OBU can carry configuration information, metadata, or coded video data.

NOTE: Tile List OBUs defined in the [AV1] specification are not supported in the current version of this specification. A future version of the specification may do so.

Temporal Units are processed by a decoder in the order given by the bitstream. Each Temporal Unit is associated with a presentation time. Some Temporal Units may contain multiple frames to be decoded but only one is presented.

NOTE: The AV1 specification defines scalability features, allowing frames from multiple layers or operating points to be present in a single temporal unit. This version of storage in ISOBMFF supports the simple storage of scalable streams in a single track, but does not specify advanced tools for handling multi-track support, layer extraction, or other scalability related use cases. A future version of the specification may do so.

Frames carried in Temporal Units may have coding dependencies on frames carried previously in the same Temporal Unit or in previous Temporal Units. Frames that can be decoded without dependencies to previous frames are of two categories: Key Frames and Intra-only Frames. Frames that cannot be decoded independently are of three categories: Inter Frames, Switch Frames, and frames with a show_existing_frame flag set to 1.

Key Frames with the show_frame flag set to 1 have the additional property that after decoding the Key Frame, all frames following the Key Frame in the bitstream can be decoded. They are called Random Access Points in [AV1].

Key Frames with the show_frame flag set to 0 are called Delayed Random Access Points. Delayed Random Access Points have the additional property that if a future Key Frame Dependent Recovery Point exists, all frames following that Key Frame Dependent Recovery Point can be decoded. A Key Frame Dependent Recovery Point is a frame with show_existing_frame set to 1 that refers to a previous Delayed Random Access Points.

2. Basic Encapsulation Scheme

This section describes the basic data structures used to signal encapsulation of AV1 bitstreams in [ISOBMFF] containers.

2.1. General Requirements & Brands

A file conformant to this specification satisfies the following:

• It SHALL conform to the normative requirements of [ISOBMFF]
• It SHALL have the av01 brand among the compatible brands array of the FileTypeBox
• It SHALL contain at least one track using an AV1SampleEntry, possibly transformed by encryption as specified in § 4 Common Encryption
• It SHOULD indicate a structural ISOBMFF brand among the compatible brands array of the FileTypeBox, such as iso6
• It MAY indicate CMAF brands as specified in § 3 CMAF AV1 track format
• It MAY indicate other brands not specified in this document provided that the associated requirements do not conflict with those given in this specification

Parsers SHALL support the structures required by the iso6 brand and MAY support structures required by further ISOBMFF structural brands.

2.2. AV1 Sample Entry

2.2.1. Definition

Sample Entry Type: av01
Container:         Sample Description Box ('stsd')
Mandatory:         Yes 
Quantity:          One or more 

2.2.2. Description

The AV1SampleEntry sample entry identifies that the track contains AV1 Samples, and uses an AV1CodecConfigurationBox.

2.2.3. Syntax

class AV1SampleEntry
extends VisualSampleEntry('av01')
{
  AV1CodecConfigurationBox config;
}

2.2.4. Semantics

Let MaxRenderWidth be the maximum RenderWidth and MaxRenderHeight be the maximum RenderHeight of all the frames in the track.

hSpacing / vSpacing = MaxRenderWidth * (max_frame_height_minus_1 + 1) /
                        ((max_frame_width_minus_1 + 1) * MaxRenderHeight)

The compressorname field of the VisualSampleEntry is an informative name. It is formatted in a fixed 32-byte field, with the first byte set to the number of bytes to be displayed, followed by that number of bytes of displayable data, followed by padding to complete 32 bytes total (including the size byte). The value "\012AOM Coding" is RECOMMENDED; the first byte is a count of the remaining bytes, here represented by \012, which (being octal 12) is decimal 10, the number of bytes in the rest of the string.

NOTE: Parsers may ignore the value of the compressorname field. It is specified in this document simply for legacy and backwards compatibility reasons.

NOTE: Multiple instances of AV1SampleEntry may be required when the track contains samples requiring a AV1CodecConfigurationBox with different characteristics.

2.3. AV1 Codec Configuration Box

2.3.1. Definition

Box Type:  av1C
Container: AV1 Sample Entry ('av01')
Mandatory: Yes 
Quantity:  Exactly One 

2.3.2. Description

2.3.3. Syntax

class AV1CodecConfigurationBox
extends Box('av1C')
{
  AV1CodecConfigurationRecord av1Config;
}

aligned(8) class AV1CodecConfigurationRecord
{
  unsigned int(1) marker = 1;
  unsigned int(7) version = 1;
  unsigned int(3) seq_profile;
  unsigned int(5) seq_level_idx_0;
  unsigned int(1) seq_tier_0;
  unsigned int(1) high_bitdepth;
  unsigned int(1) twelve_bit;
  unsigned int(1) monochrome;
  unsigned int(1) chroma_subsampling_x;
  unsigned int(1) chroma_subsampling_y;
  unsigned int(2) chroma_sample_position;
  unsigned int(3) reserved = 0;

  unsigned int(1) initial_presentation_delay_present;
  if(initial_presentation_delay_present) {
    unsigned int(4) initial_presentation_delay_minus_one;
  } else {
    unsigned int(4) reserved = 0;
  }

  unsigned int(8) configOBUs[];
}

2.3.4. Semantics

NOTE: The marker bit ensures that the bit pattern of the first byte of the AV1CodecConfigurationRecord cannot be mistaken for an OBU Header byte.

The version field indicates the version of the AV1CodecConfigurationRecord. The value SHALL be set to 1 for AV1CodecConfigurationRecord.

The initial_presentation_delay_present field indicates the presence of the initial_presentation_delay_minus_one field.

The initial_presentation_delay_minus_one field indicates the number of samples (minus one) that need to be decoded prior to starting the presentation of the first sample associated with this sample entry in order to guarantee that each sample will be decoded prior to its presentation time under the constraints of the first level value indicated by seq_level_idx in the Sequence Header OBU (in the configOBUs field or in the associated samples). More precisely, the following procedure SHALL not return any error:

• construct a hypothetical bitstream consisting of the OBUs carried in the sample entry followed by the OBUs carried in all the samples referring to that sample entry,

• set the first initial_display_delay_minus_1 field of each Sequence Header OBU to the number of frames minus one contained in the first initial_presentation_delay_minus_one + 1 samples,

• set the frame_presentation_time field of the frame header of each presentable frame such that it matches the presentation time difference between the sample carrying this frame and the previous sample (if it exists, 0 otherwise),

• apply the decoder model specified in [AV1] to this hypothetical bitstream using the first operating point. If buffer_removal_time information is present in bitstream for this operating point, the decoding schedule mode SHALL be applied, otherwise the resource availability mode SHALL be applied.

NOTE: With the above procedure, when smooth presentation can be guaranteed after decoding the first sample, initial_presentation_delay_minus_one is 0.

NOTE: Because the above procedure considers all OBUs in all samples associated with a sample entry, if these OBUS form multiple coded video sequences which would have different values of initial_presentation_delay_minus_one if considered separately, the sample entry would signal the larger value.

a b c d e f g h

[a] [b c d] [e] [f] [g] [h]

[a] [b] [c] [d e f] [g] [h]

• From any sync sample, an AV1 bitstream is formed by first outputting the OBUs contained in the AV1CodecConfigurationBox and then by outputing all OBUs in the samples themselves, in order, starting from the sync sample.

NOTE: If a Sequence Header OBU is stored in the configOBUs field, because a sync sample also stores the Sequence Header OBU, this procedure will produce 2 consecutive and identical Sequence Header OBUs. Compliant [AV1] decoders are expected to handle that.

• From any sample marked with the AV1ForwardKeyFrameSampleGroupEntry, an AV1 bitstream is formed by first outputting the OBUs contained in the AV1CodecConfigurationBox and then by outputing all OBUs in the sample itself, then by outputting all OBUs in the samples, in order, starting from the sample at the distance indicated by the sample group.

Additionally, the configOBUs field SHALL contain at most one Sequence Header OBU and if present, it SHALL be the first OBU.

NOTE: The configOBUs field is expected to contain only one Sequence Header OBU and zero or more Metadata OBUs when applicable to all the associated samples.

OBUs stored in the configOBUs field follow the open_bitstream_unit Low Overhead Bitstream Format syntax as specified in [AV1]. The flag obu_has_size_field SHALL be set to 1, indicating that the size of the OBU payload follows the header, and that it is coded using LEB128.

The presentation times of AV1 samples are given by the ISOBMFF structures. The timing_info_present_flag in the Sequence Header OBU (in the configOBUs field or in the associated samples) SHOULD be set to 0. If set to 1, the timing_info structure of the Sequence Header OBU, the frame_presentation_time and buffer_removal_time fields of the Frame Header OBUs, if present, SHALL be ignored for the purpose of timed processing of the ISOBMFF file.

• If any of colour_primaries, transfer_characteristics, and matrix_coefficients is specified with a value other than 2 in the Sequence Header OBU (in the configOBUs field or in the associated samples), then the value of that field in the colr box SHALL match the value in the Sequence Header OBU.
• If any of colour_primaries, transfer_characteristics, and matrix_coefficients is NOT specified or is specified with a value equal to 2 in the Sequence Header OBU, then the value of that field in the colr box MAY be different and the value in the colr box overrides the value in the bitstream, as specified in the ColourInformationBox definition in ISOBMFF.
• The value of full_range_flag in the colr box SHALL match the color_range flag in the Sequence Header OBU.

For sample entries corresponding to HDR content, the following applies:

• The ContentLightLevelBox clli SHOULD be present. Its values SHALL equal the values contained in the Metadata OBU of type METADATA_TYPE_HDR_CLL, if present (in the configOBUs or in the samples).
• The MasteringDisplayColourVolumeBox mdcv SHOULD be present. Its decoded values SHALL equal (except for precision issues, see the second note below) the decoded values contained in the Metadata OBU of type METADATA_TYPE_HDR_MDCV, if present (in the configOBUs or in the samples).

NOTE: The MasteringDisplayColourVolumeBox mdcv and ContentLightLevelBox clli have identical syntax to the SMPTE2086MasteringDisplayMetadataBox SmDm and ContentLightLevelBox CoLL, except that they are of type Box and not FullBox. However, the semantics of the MasteringDisplayColourVolumeBox and SMPTE2086MasteringDisplayMetadataBox have important differences and SHOULD not be confused.

NOTE: In the MasteringDisplayColourVolumeBox mdcv, the representation and coding of some fields are different from the representation and coding of the corresponding fields in the Metadata OBU of type METADATA_TYPE_HDR_MDCV. Because of these differences, some values coded in the MasteringDisplayColourVolumeBox mdcv may not have the same precision in the Metadata OBU of type METADATA_TYPE_HDR_MDCV, and vice versa.

2.4. AV1 Sample Format

For tracks using the AV1SampleEntry, an AV1 Sample has the following constraints:

• the sample data SHALL be a sequence of OBUs forming a Temporal Unit,
• each OBU SHALL follow the open_bitstream_unit Low Overhead Bitstream Format syntax as specified in [AV1]. Each OBU SHALL have the obu_has_size_field set to 1 except for the last OBU in the sample, for which obu_has_size_field MAY be set to 0, in which case it is assumed to fill the remainder of the sample,

NOTE: When extracting OBUs from an ISOBMFF file, and depending on the capabilities of the decoder processing these OBUs, ISOBMFF parsers MAY need to either: set the obu_has_size_field to 1 for some OBUs if not already set, add the size field in this case, and add Temporal Delimiter OBU; or use the length-delimited bitstream format as defined in Annex B of AV1. If encryption is used, similar operations MAY have to be done before or after decryption depending on the demuxer/decryptor/decoder architecture.

• OBU trailing bits SHOULD be limited to byte alignment and SHOULD not be used for padding,
• OBUs of type OBU_TEMPORAL_DELIMITER, OBU_PADDING, or OBU_REDUNDANT_FRAME_HEADER SHOULD NOT be used.
• OBUs of type OBU_TILE_LIST SHALL NOT be used.

• Its first frame is a Key Frame that has show_frame flag set to 1,

• It contains a Sequence Header OBU before the first Frame Header OBU.

NOTE: Within this definition, a sync sample may contain additional frames that are not Key Frames. The fact that none of them is the first frame in the temporal unit ensures that they are decodable.

NOTE: Other types of OBUs such as metadata OBUs could be present before the Sequence Header OBU.

Additionally, if a file contains multiple tracks that are alternative representations of the same content, in particular using Switch Frames, those tracks SHOULD be marked as belonging to the same alternate group and SHOULD use a track selection box with an appropriate attribute (e.g. bitr).

Unless explicitely stated, the grouping_type_parameter is not defined for the SampleToGroupBox with grouping types defined in this specification.

2.5. AV1 Forward Key Frame sample group entry

2.5.1. Definition

Group Type: av1f
Container:  Sample Group Description Box ('sgpd')
Mandatory:  No 
Quantity:   Zero or more .

2.5.2. Description

The AV1ForwardKeyFrameSampleGroupEntry documents samples that contain a Delayed Random Access Point that are followed at a given distance in the bitstream by a Key Frame Dependent Recovery Point.

2.5.3. Syntax

class AV1ForwardKeyFrameSampleGroupEntry
extends VisualSampleGroupEntry('av1f')
{
  unsigned int(8) fwd_distance;
}

2.5.4. Semantics

The fwd_distance field indicates the number of samples between this sample and the next sample containing the associated Key Frame Dependent Recovery Point. 0 means the next sample.

2.6. AV1 Multi-Frame sample group entry

2.6.1. Definition

Group Type: av1m
Container:  Sample Group Description Box ('sgpd')
Mandatory:  No 
Quantity:   Zero or more .

2.6.2. Description

The AV1MultiFrameSampleGroupEntry documents samples that contain multiple frames.

2.6.3. Syntax

class AV1MultiFrameSampleGroupEntry
extends VisualSampleGroupEntry('av1m')
{
}

2.7. AV1 Switch Frame sample group entry

2.7.1. Definition

Group Type: av1s
Container:  Sample Group Description Box ('sgpd')
Mandatory:  No 
Quantity:   Zero or more .

2.7.2. Description

The AV1SwitchFrameSampleGroupEntry documents samples that start with a Switch Frame.

2.7.3. Syntax

class AV1SwitchFrameSampleGroupEntry
extends VisualSampleGroupEntry('av1s')
{
}

2.8. AV1 Metadata sample group entry

2.8.1. Definition

Group Type: av1M
Container:  Sample Group Description Box ('sgpd')
Mandatory:  No 
Quantity:   Zero or more .

2.8.2. Description

The AV1MetadataSampleGroupEntry documents samples that contain metadata OBUs. The grouping_type_parameter can be used to identify samples containing metadata OBUs of a given type. If no grouping_type_parameter is provided, the sample group entry identifies samples containing metadata OBUs for which the metadata_type is unknown.

2.8.3. Syntax

class AV1MetadataSampleGroupEntry
extends VisualSampleGroupEntry('av1M')
{
}

For this sample group entry, the grouping_type_parameter syntax is as follows:

{
   unsigned int (8) metadata_type;
   unsigned int (24) metadata_specific_parameters;
}

2.8.4. Semantics

metadata_type is a 8-bit field whose value is the value of the metadata_type field defined in [AV1], when it is equal or lower than 255. metadata_type values above 255 are not supported by this sample group.

metadata_specific_parameters provides an additional part of the grouping_type_parameter, which MAY be used to distinguish different sample groups having the same metadata_type but different sub-parameters. In this version of the specification, metadata_specific_parameters is only defined when metadata_type is set to METADATA_TYPE_ITUT_T35 in which case its value SHALL be set to the first 24 bits of the metadata_itut_t35 structure. For other types of metadata, its value SHOULD be set to 0. In all cases, when the grouping_type_parameter is used, readers processing sample groups SHALL use the entire value of grouping_type_parameter.

NOTE: The current design is constrained to using only 3 bytes for ITU-T T.35 identification, which typically suffices for the itu_t_t35_country_code and the itu_t_t35_terminal_provider_code identification. Consequently, inspection of the payload becomes essential if an implementation aims to determine the values of subsequent bytes, for instance, the itu_t_t35_terminal_provider_oriented_code.

3. CMAF AV1 track format

[CMAF] defines structural constraints on ISOBMFF files additional to [ISOBMFF] for the purpose of, for example, adaptive streaming or for protected files. Conformance to these structural constraints is signaled by the presence of the brands cmfc or cmf2 in the FileTypeBox.

NOTE: It is important to note that while the cmf2 brand does not introduce any new requirements compared to the previously defined cmfc brand, it further restricts the cmfc constraints.

For compatibility reasons, when the cmf2 brand is used, the cmfc brand SHOULD also be included in the FileTypeBox.

If a CMAF Video Track uses the brand av01, it is called a CMAF AV1 Track and the following constraints, defining the CMAF Media Profile for AV1, apply:

• it SHALL use an AV1SampleEntry, possibly transformed by encryption as specified in § 4 Common Encryption
• it MAY use multiple sample entries, and in that case the following values SHALL not change in the track:

  • seq_profile

  • still_picture

  • the first value of seq_level_idx

  • the first value of seq_tier

  • color_config

  • initial_presentation_delay_minus_one

4. Common Encryption

• crypt_byte_block = 1 and skip_byte_block = 9,
• crypt_byte_block = 5 and skip_byte_block = 5,
• crypt_byte_block = 10 and skip_byte_block = 0.

4.1. General Subsample Encryption constraints

Within protected samples, the following constraints apply:

• Protected samples SHALL be exactly spanned by one or more contiguous subsamples.
  • An OBU MAY be spanned by one or more subsamples, especially when it has multiple ranges of protected data. However, as recommended in [CENC], subsample encryption entries SHOULD be as compactly represented as possible. For example, only one subsample encryption entry should be created for multiple consecutive unprotected OBUs as well as the first unprotected and protected parts of the following protected OBU, if such protected OBU exists.
  • A large unprotected OBU whose data size is larger than the maximum size of a single BytesOfClearData field MAY be spanned by multiple subsamples with zero size BytesOfProtectedData.
• All OBU Headers and associated obu_size fields SHALL be unprotected.
• Temporal Delimiter OBUs, Sequence Header OBUs, and Frame Header OBUs (including within a Frame OBU), Redundant Frame Header OBUs and Padding OBUs SHALL be unprotected.
• Metadata OBUs SHALL NOT be protected.
• Tile Group OBUs and Frame OBUs SHALL be protected. Within Tile Group OBUs or Frame OBUs, the following applies:

  • for the protected scheme cbcs:

    • BytesOfProtectedData SHALL start on the first byte and end on the last byte of the decode_tile structure (including any trailing bits).

    NOTE: As a result of the above, BytesOfProtectedData is not necessarily a multiple of 16 bytes and partial blocks are included in the BytesOfProtectedData range, but per [CENC] the data is not be encrypted.

    • A subsample SHALL be created for each tile, even if its size is less than 16 bytes.
    • All other parts of Tile Group OBUs and Frame OBUs SHALL be unprotected .

  • for the protected scheme cenc:

    • BytesOfProtectedData SHALL be a multiple of 16 bytes.
    • BytesOfProtectedData SHALL end on the last byte of the decode_tile structure (including any trailing bits).
    • BytesOfProtectedData SHALL span all complete 16-byte blocks of the decode_tile structure (including any trailing bits).

    NOTE: As a result of the above, partial blocks are not used and it is possible that BytesOfProtectedData does not start at the first byte of the decode_tile structure, but some number of bytes following that.

    • A subsample SHALL be created for each tile that has a decode_tile structure whose size (including any trailing bits) is larger or equal to 16 bytes. If it is less than 16 bytes, per the rules above, the decode_tile structure is not encrypted and the corresponding bytes SHOULD be included in the BytesOfClearData field of a surrounding subsample, if any.
    • All other parts of Tile Group OBUs and Frame OBUs SHALL be unprotected .

NOTE: Version 0 of the saiz box as defined in [ISOBMFF] can only describe Sample Auxiliary Information whose size does not exceed 255 bytes. In the case of AV1 streams that use a large number of tiles or a large number of frames per sample, when following the encryption requirements defined in this section the size of the corresponding Sample Auxiliary Information may exceed that limit. Until updates are made to [ISOBMFF], authors are encouraged to avoid reaching the limit.

4.2. Subsample Encryption Illustration

5. Codecs Parameter String

DASH and other applications require defined values for the Codecs parameter specified in [RFC6381] for ISO Media tracks. The codecs parameter string for the AOM AV1 codec is as follows:

<sample entry 4CC>.<profile>.<level><tier>.<bitDepth>.<monochrome>.<chromaSubsampling>.
<colorPrimaries>.<transferCharacteristics>.<matrixCoefficients>.<videoFullRangeFlag>

All fields following the sample entry 4CC are expressed as double digit decimals, unless indicated otherwise. Leading or trailing zeros cannot be omitted.

The colorPrimaries, transferCharacteristics, matrixCoefficients, and videoFullRangeFlag parameter values are set as follows:

• If a colr box with colour_type set to nclx is present, the colorPrimaries, transferCharacteristics, matrixCoefficients, and videoFullRangeFlag parameter values SHALL equal the values of matching fields in the colr box.

• Otherwise, a colr box with colour_type set to nclx is absent.

  • The videoFullRangeFlag parameter value SHALL equal the color_range flag in the Sequence Header OBU.
  • If the color_description_present_flag is set to 1 in the Sequence Header OBU, the colorPrimaries, transferCharacteristics, and matrixCoefficients parameter values SHALL equal the values of matching fields in the Sequence Header OBU.

  • Otherwise, the color_description_present_flag is set to 0 in the Sequence Header OBU. The colorPrimaries, transferCharacteristics, and matrixCoefficients parameter values SHOULD be set to the default values below.

NOTE: It is encouraged that a colr box be present. The colr box simplifies the task of the codecs parameter generator.

NOTE: If the encoded video uses a color space that is not registered in [CICP] and the file doesn’t explicitly code 2/2/2 either in the colr box or in the Sequence Header OBU (with color_description_present_flag set to 1), the derived codecs parameter string will declare ITU-R BT.709 (1/1/1). To avoid this behavior, a colr box should be used.

The videoFullRangeFlag is represented by a single digit.

For example, codecs="av01.0.04M.10.0.112.09.16.09.0" represents AV1 Main Profile, level 3.0, Main tier, 10-bit content, non-monochrome, with 4:2:0 chroma subsampling co-located with (0, 0) luma sample, ITU-R BT.2100 color primaries, ITU-R BT.2100 PQ transfer characteristics, ITU-R BT.2100 YCbCr color matrix, and studio swing representation.

The string codecs="av01.0.01M.08" in this case would represent AV1 Main Profile, level 2.1, Main tier, 8-bit content with 4:2:0 chroma subsampling, ITU-R BT.709 color primaries, transfer characteristics, matrix coefficients, and studio swing representation.

If any character that is not '.', digits, part of the AV1 4CC, or a tier value is encountered, the string SHALL be interpreted ignoring all the characters starting from that character.

6. Changes since v1.2.0 release

• Clarify the color info in the codecs string.

• Update encryption diagram.

• Remove clap restriction.

• Clarify compactness and encryption.

• Add a NOTE to clarify the mdcv box.

• Clarify requirements on sample entry when encryption is used.

• Clarify the colr box.

• Update the CMAF section to mention the cmf2 brand.

• Make assert-id’s unique and add them to the previous version.

• Correct the page numbers in links to the AV1 spec.

• Shorten long lines

• Add a note on the identification of the T.35 message.

• Clarify storage of Sequence Header OBU.

• Use undated reference for ISOBMFF.

• Add note for encryption description limit.

• Add a link to the conformance page.