opus/doc/draft-ietf-codec-oggopus.xml

<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE rfc SYSTEM 'rfc2629.dtd' [
<!ENTITY rfc2119 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.2119.xml'>
<!ENTITY rfc3533 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3533.xml'>
<!ENTITY rfc3629 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.3629.xml'>
<!ENTITY rfc4732 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.4732.xml'>
<!ENTITY rfc5334 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.5334.xml'>
<!ENTITY rfc6381 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.6381.xml'>
<!ENTITY rfc6716 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.6716.xml'>
<!ENTITY rfc6982 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.6982.xml'>
<!ENTITY rfc7587 PUBLIC '' 'http://xml.resource.org/public/rfc/bibxml/reference.RFC.7587.xml'>
]>
<?rfc toc="yes" symrefs="yes" ?>

<rfc ipr="trust200902" category="std" docName="draft-ietf-codec-oggopus-09">

<front>
<title abbrev="Ogg Opus">Ogg Encapsulation for the Opus Audio Codec</title>
<author initials="T.B." surname="Terriberry" fullname="Timothy B. Terriberry">
<organization>Mozilla Corporation</organization>
<address>
<postal>
<street>650 Castro Street</street>
<city>Mountain View</city>
<region>CA</region>
<code>94041</code>
<country>USA</country>
</postal>
<phone>+1 650 903-0800</phone>
<email>tterribe@xiph.org</email>
</address>
</author>

<author initials="R." surname="Lee" fullname="Ron Lee">
<organization>Voicetronix</organization>
<address>
<postal>
<street>246 Pulteney Street, Level 1</street>
<city>Adelaide</city>
<region>SA</region>
<code>5000</code>
<country>Australia</country>
</postal>
<phone>+61 8 8232 9112</phone>
<email>ron@debian.org</email>
</address>
</author>

<author initials="R." surname="Giles" fullname="Ralph Giles">
<organization>Mozilla Corporation</organization>
<address>
<postal>
<street>163 West Hastings Street</street>
<city>Vancouver</city>
<region>BC</region>
<code>V6B 1H5</code>
<country>Canada</country>
</postal>
<phone>+1 778 785 1540</phone>
<email>giles@xiph.org</email>
</address>
</author>

<date day="23" month="November" year="2015"/>
<area>RAI</area>
<workgroup>codec</workgroup>

<abstract>
<t>
This document defines the Ogg encapsulation for the Opus interactive speech and
 audio codec.
This allows data encoded in the Opus format to be stored in an Ogg logical
 bitstream.
</t>
</abstract>
</front>

<middle>
<section anchor="intro" title="Introduction">
<t>
The IETF Opus codec is a low-latency audio codec optimized for both voice and
 general-purpose audio.
See <xref target="RFC6716"/> for technical details.
This document defines the encapsulation of Opus in a continuous, logical Ogg
 bitstream&nbsp;<xref target="RFC3533"/>.
Ogg encapsulation provides Opus with a long-term storage format supporting
 all of the essential features, including metadata, fast and accurate seeking,
 corruption detection, recapture after errors, low overhead, and the ability to
 multiplex Opus with other codecs (including video) with minimal buffering.
It also provides a live streamable format, capable of delivery over a reliable
 stream-oriented transport, without requiring all the data, or even the total
 length of the data, up-front, in a form that is identical to the on-disk
 storage format.
</t>
<t>
Ogg bitstreams are made up of a series of 'pages', each of which contains data
 from one or more 'packets'.
Pages are the fundamental unit of multiplexing in an Ogg stream.
Each page is associated with a particular logical stream and contains a capture
 pattern and checksum, flags to mark the beginning and end of the logical
 stream, and a 'granule position' that represents an absolute position in the
 stream, to aid seeking.
A single page can contain up to 65,025 octets of packet data from up to 255
 different packets.
Packets can be split arbitrarily across pages, and continued from one page to
 the next (allowing packets much larger than would fit on a single page).
Each page contains 'lacing values' that indicate how the data is partitioned
 into packets, allowing a demuxer to recover the packet boundaries without
 examining the encoded data.
A packet is said to 'complete' on a page when the page contains the final
 lacing value corresponding to that packet.
</t>
<t>
This encapsulation defines the contents of the packet data, including
 the necessary headers, the organization of those packets into a logical
 stream, and the interpretation of the codec-specific granule position field.
It does not attempt to describe or specify the existing Ogg container format.
Readers unfamiliar with the basic concepts mentioned above are encouraged to
 review the details in <xref target="RFC3533"/>.
</t>

</section>

<section anchor="terminology" title="Terminology">
<t>
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 <xref target="RFC2119"/>.
</t>

<t>
Implementations that fail to satisfy one or more "MUST" requirements are
 considered non-compliant.
Implementations that satisfy all "MUST" requirements, but fail to satisfy one
 or more "SHOULD" requirements are said to be "conditionally compliant".
All other implementations are "unconditionally compliant".
</t>

</section>

<section anchor="packet_organization" title="Packet Organization">
<t>
An Ogg Opus stream is organized as follows.
</t>
<t>
There are two mandatory header packets.
The first packet in the logical Ogg bitstream MUST contain the identification
 (ID) header, which uniquely identifies a stream as Opus audio.
The format of this header is defined in <xref target="id_header"/>.
It MUST be placed alone (without any other packet data) on the first page of
 the logical Ogg bitstream, and MUST complete on that page.
This page MUST have its 'beginning of stream' flag set.
</t>
<t>
The second packet in the logical Ogg bitstream MUST contain the comment header,
 which contains user-supplied metadata.
The format of this header is defined in <xref target="comment_header"/>.
It MAY span multiple pages, beginning on the second page of the logical
 stream.
However many pages it spans, the comment header packet MUST finish the page on
 which it completes.
</t>
<t>
All subsequent pages are audio data pages, and the Ogg packets they contain are
 audio data packets.
Each audio data packet contains one Opus packet for each of N different
 streams, where N is typically one for mono or stereo, but MAY be greater than
 one for multichannel audio.
The value N is specified in the ID header (see
 <xref target="channel_mapping"/>), and is fixed over the entire length of the
 logical Ogg bitstream.
</t>
<t>
The first (N&nbsp;-&nbsp;1) Opus packets, if any, are packed one after another
 into the Ogg packet, using the self-delimiting framing from Appendix&nbsp;B of
 <xref target="RFC6716"/>.
The remaining Opus packet is packed at the end of the Ogg packet using the
 regular, undelimited framing from Section&nbsp;3 of <xref target="RFC6716"/>.
All of the Opus packets in a single Ogg packet MUST be constrained to have the
 same duration.
An implementation of this specification SHOULD treat any Opus packet whose
 duration is different from that of the first Opus packet in an Ogg packet as
 if it were a malformed Opus packet with an invalid TOC sequence.
</t>
<t>
The coding mode (SILK, Hybrid, or CELT), audio bandwidth, channel count,
 duration (frame size), and number of frames per packet, are indicated in the
 TOC (table of contents) sequence at the beginning of each Opus packet, as
 described in Section&nbsp;3.1 of&nbsp;<xref target="RFC6716"/>.
The combination of mode, audio bandwidth, and frame size is referred to as
 the configuration of an Opus packet.
</t>
<t>
The first audio data page SHOULD NOT have the 'continued packet' flag set
 (which would indicate the first audio data packet is continued from a previous
 page).
Packets MUST be placed into Ogg pages in order until the end of stream.
Audio packets MAY span page boundaries.
An implementation MUST treat a zero-octet audio data packet as if it were a
 malformed Opus packet as described in
 Section&nbsp;3.4 of&nbsp;<xref target="RFC6716"/>.
</t>
<t>
The last page SHOULD have the 'end of stream' flag set, but implementations
 need to be prepared to deal with truncated streams that do not have a page
 marked 'end of stream'.
The final packet on the last page SHOULD NOT be a continued packet, i.e., the
 final lacing value SHOULD be less than 255.
There MUST NOT be any more pages in an Opus logical bitstream after a page
 marked 'end of stream'.
</t>
</section>

<section anchor="granpos" title="Granule Position">
<t>
The granule position MUST be zero for the ID header page and the
 page where the comment header completes.
That is, the first page in the logical stream, and the last header
 page before the first audio data page both have a granule position of zero.
</t>
<t>
The granule position of an audio data page encodes the total number of PCM
 samples in the stream up to and including the last fully-decodable sample from
 the last packet completed on that page.
The granule position of the first audio data page will usually be larger than
 zero, as described in <xref target="start_granpos_restrictions"/>.
</t>

<t>
A page that is entirely spanned by a single packet (that completes on a
 subsequent page) has no granule position, and the granule position field MUST
 be set to the special value '-1' in two's complement.
</t>

<t>
The granule position of an audio data page is in units of PCM audio samples at
 a fixed rate of 48&nbsp;kHz (per channel; a stereo stream's granule position
 does not increment at twice the speed of a mono stream).
It is possible to run an Opus decoder at other sampling rates, but the value
 in the granule position field always counts samples assuming a 48&nbsp;kHz
 decoding rate, and the rest of this specification makes the same assumption.
</t>

<t>
The duration of an Opus packet can be any multiple of 2.5&nbsp;ms, up to a
 maximum of 120&nbsp;ms.
This duration is encoded in the TOC sequence at the beginning of each packet.
The number of samples returned by a decoder corresponds to this duration
 exactly, even for the first few packets.
For example, a 20&nbsp;ms packet fed to a decoder running at 48&nbsp;kHz will
 always return 960&nbsp;samples.
A demuxer can parse the TOC sequence at the beginning of each Ogg packet to
 work backwards or forwards from a packet with a known granule position (i.e.,
 the last packet completed on some page) in order to assign granule positions
 to every packet, or even every individual sample.
The one exception is the last page in the stream, as described below.
</t>

<t>
All other pages with completed packets after the first MUST have a granule
 position equal to the number of samples contained in packets that complete on
 that page plus the granule position of the most recent page with completed
 packets.
This guarantees that a demuxer can assign individual packets the same granule
 position when working forwards as when working backwards.
For this to work, there cannot be any gaps.
</t>

<section anchor="gap-repair" title="Repairing Gaps in Real-time Streams">
<t>
In order to support capturing a real-time stream that has lost or not
 transmitted packets, a muxer SHOULD emit packets that explicitly request the
 use of Packet Loss Concealment (PLC) in place of the missing packets.
Implementations that fail to do so still MUST NOT increment the granule
 position for a page by anything other than the number of samples contained in
 packets that actually complete on that page.
</t>
<t>
Only gaps that are a multiple of 2.5&nbsp;ms are repairable, as these are the
 only durations that can be created by packet loss or discontinuous
 transmission.
Muxers need not handle other gap sizes.
Creating the necessary packets involves synthesizing a TOC byte (defined in
Section&nbsp;3.1 of&nbsp;<xref target="RFC6716"/>)&mdash;and whatever
 additional internal framing is needed&mdash;to indicate the packet duration
 for each stream.
The actual length of each missing Opus frame inside the packet is zero bytes,
 as defined in Section&nbsp;3.2.1 of&nbsp;<xref target="RFC6716"/>.
</t>

<t>
Zero-byte frames MAY be packed into packets using any of codes&nbsp;0, 1,
 2, or&nbsp;3.
When successive frames have the same configuration, the higher code packings
 reduce overhead.
Likewise, if the TOC configuration matches, the muxer MAY further combine the
 empty frames with previous or subsequent non-zero-length frames (using
 code&nbsp;2 or VBR code&nbsp;3).
</t>

<t>
<xref target="RFC6716"/> does not impose any requirements on the PLC, but this
 section outlines choices that are expected to have a positive influence on
 most PLC implementations, including the reference implementation.
Synthesized TOC sequences SHOULD maintain the same mode, audio bandwidth,
 channel count, and frame size as the previous packet (if any).
This is the simplest and usually the most well-tested case for the PLC to
 handle and it covers all losses that do not include a configuration switch,
 as defined in Section&nbsp;4.5 of&nbsp;<xref target="RFC6716"/>.
</t>

<t>
When a previous packet is available, keeping the audio bandwidth and channel
 count the same allows the PLC to provide maximum continuity in the concealment
 data it generates.
However, if the size of the gap is not a multiple of the most recent frame
 size, then the frame size will have to change for at least some frames.
Such changes SHOULD be delayed as long as possible to simplify
 things for PLC implementations.
</t>

<t>
As an example, a 95&nbsp;ms gap could be encoded as nineteen 5&nbsp;ms frames
 in two bytes with a single CBR code&nbsp;3 packet.
If the previous frame size was 20&nbsp;ms, using four 20&nbsp;ms frames
 followed by three 5&nbsp;ms frames requires 4&nbsp;bytes (plus an extra byte
 of Ogg lacing overhead), but allows the PLC to use its well-tested steady
 state behavior for as long as possible.
The total bitrate of the latter approach, including Ogg overhead, is about
 0.4&nbsp;kbps, so the impact on file size is minimal.
</t>

<t>
Changing modes is discouraged, since this causes some decoder implementations
 to reset their PLC state.
However, SILK and Hybrid mode frames cannot fill gaps that are not a multiple
 of 10&nbsp;ms.
If switching to CELT mode is needed to match the gap size, a muxer SHOULD do
 so at the end of the gap to allow the PLC to function for as long as possible.
</t>

<t>
In the example above, if the previous frame was a 20&nbsp;ms SILK mode frame,
 the better solution is to synthesize a packet describing four 20&nbsp;ms SILK
 frames, followed by a packet with a single 10&nbsp;ms SILK
 frame, and finally a packet with a 5&nbsp;ms CELT frame, to fill the 95&nbsp;ms
 gap.
This also requires four bytes to describe the synthesized packet data (two
 bytes for a CBR code 3 and one byte each for two code 0 packets) but three
 bytes of Ogg lacing overhead are needed to mark the packet boundaries.
At 0.6 kbps, this is still a minimal bitrate impact over a naive, low quality
 solution.
</t>

<t>
Since medium-band audio is an option only in the SILK mode, wideband frames
 SHOULD be generated if switching from that configuration to CELT mode, to
 ensure that any PLC implementation which does try to migrate state between
 the modes will be able to preserve all of the available audio bandwidth.
</t>

</section>

<section anchor="preskip" title="Pre-skip">
<t>
There is some amount of latency introduced during the decoding process, to
 allow for overlap in the CELT mode, stereo mixing in the SILK mode, and
 resampling.
The encoder might have introduced additional latency through its own resampling
 and analysis (though the exact amount is not specified).
Therefore, the first few samples produced by the decoder do not correspond to
 real input audio, but are instead composed of padding inserted by the encoder
 to compensate for this latency.
These samples need to be stored and decoded, as Opus is an asymptotically
 convergent predictive codec, meaning the decoded contents of each frame depend
 on the recent history of decoder inputs.
However, a player will want to skip these samples after decoding them.
</t>

<t>
A 'pre-skip' field in the ID header (see <xref target="id_header"/>) signals
 the number of samples which SHOULD be skipped (decoded but discarded) at the
 beginning of the stream.
This amount need not be a multiple of 2.5&nbsp;ms, MAY be smaller than a single
 packet, or MAY span the contents of several packets.
These samples are not valid audio, and SHOULD NOT be played.
</t>

<t>
For example, if the first Opus frame uses the CELT mode, it will always
 produce 120 samples of windowed overlap-add data.
However, the overlap data is initially all zeros (since there is no prior
 frame), meaning this cannot, in general, accurately represent the original
 audio.
The SILK mode requires additional delay to account for its analysis and
 resampling latency.
The encoder delays the original audio to avoid this problem.
</t>

<t>
The pre-skip field MAY also be used to perform sample-accurate cropping of
 already encoded streams.
In this case, a value of at least 3840&nbsp;samples (80&nbsp;ms) provides
 sufficient history to the decoder that it will have converged
 before the stream's output begins.
</t>

</section>

<section anchor="pcm_sample_position" title="PCM Sample Position">
<t>
The PCM sample position is determined from the granule position using the
 formula
</t>
<figure align="center">
<artwork align="center"><![CDATA[
'PCM sample position' = 'granule position' - 'pre-skip' .
]]></artwork>
</figure>

<t>
For example, if the granule position of the first audio data page is 59,971,
 and the pre-skip is 11,971, then the PCM sample position of the last decoded
 sample from that page is 48,000.
</t>
<t>
This can be converted into a playback time using the formula
</t>
<figure align="center">
<artwork align="center"><![CDATA[
                  'PCM sample position'
'playback time' = --------------------- .
                         48000.0
]]></artwork>
</figure>

<t>
The initial PCM sample position before any samples are played is normally '0'.
In this case, the PCM sample position of the first audio sample to be played
 starts at '1', because it marks the time on the clock
 <spanx style="emph">after</spanx> that sample has been played, and a stream
 that is exactly one second long has a final PCM sample position of '48000',
 as in the example here.
</t>

<t>
Vorbis streams use a granule position smaller than the number of audio samples
 contained in the first audio data page to indicate that some of those samples
 are trimmed from the output (see <xref target="vorbis-trim"/>).
However, to do so, Vorbis requires that the first audio data page contains
 exactly two packets, in order to allow the decoder to perform PCM position
 adjustments before needing to return any PCM data.
Opus uses the pre-skip mechanism for this purpose instead, since the encoder
 might introduce more than a single packet's worth of latency, and since very
 large packets in streams with a very large number of channels might not fit
 on a single page.
</t>
</section>

<section anchor="end_trimming" title="End Trimming">
<t>
The page with the 'end of stream' flag set MAY have a granule position that
 indicates the page contains less audio data than would normally be returned by
 decoding up through the final packet.
This is used to end the stream somewhere other than an even frame boundary.
The granule position of the most recent audio data page with completed packets
 is used to make this determination, or '0' is used if there were no previous
 audio data pages with a completed packet.
The difference between these granule positions indicates how many samples to
 keep after decoding the packets that completed on the final page.
The remaining samples are discarded.
The number of discarded samples SHOULD be no larger than the number decoded
 from the last packet.
</t>
</section>

<section anchor="start_granpos_restrictions"
 title="Restrictions on the Initial Granule Position">
<t>
The granule position of the first audio data page with a completed packet MAY
 be larger than the number of samples contained in packets that complete on
 that page, however it MUST NOT be smaller, unless that page has the 'end of
 stream' flag set.
Allowing a granule position larger than the number of samples allows the
 beginning of a stream to be cropped or a live stream to be joined without
 rewriting the granule position of all the remaining pages.
This means that the PCM sample position just before the first sample to be
 played MAY be larger than '0'.
Synchronization when multiplexing with other logical streams still uses the PCM
 sample position relative to '0' to compute sample times.
This does not affect the behavior of pre-skip: exactly 'pre-skip' samples
 SHOULD be skipped from the beginning of the decoded output, even if the
 initial PCM sample position is greater than zero.
</t>

<t>
On the other hand, a granule position that is smaller than the number of
 decoded samples prevents a demuxer from working backwards to assign each
 packet or each individual sample a valid granule position, since granule
 positions are non-negative.
An implementation MUST reject as invalid any stream where the granule position
 is smaller than the number of samples contained in packets that complete on
 the first audio data page with a completed packet, unless that page has the
 'end of stream' flag set.
It MAY defer this action until it decodes the last packet completed on that
 page.
</t>

<t>
If that page has the 'end of stream' flag set, a demuxer MUST reject as invalid
 any stream where its granule position is smaller than the 'pre-skip' amount.
This would indicate that there are more samples to be skipped from the initial
 decoded output than exist in the stream.
If the granule position is smaller than the number of decoded samples produced
 by the packets that complete on that page, then a demuxer MUST use an initial
 granule position of '0', and can work forwards from '0' to timestamp
 individual packets.
If the granule position is larger than the number of decoded samples available,
 then the demuxer MUST still work backwards as described above, even if the
 'end of stream' flag is set, to determine the initial granule position, and
 thus the initial PCM sample position.
Both of these will be greater than '0' in this case.
</t>
</section>

<section anchor="seeking_and_preroll" title="Seeking and Pre-roll">
<t>
Seeking in Ogg files is best performed using a bisection search for a page
 whose granule position corresponds to a PCM position at or before the seek
 target.
With appropriately weighted bisection, accurate seeking can be performed with
 just three or four bisections even in multi-gigabyte files.
See <xref target="seeking"/> for general implementation guidance.
</t>

<t>
When seeking within an Ogg Opus stream, an implementation SHOULD start decoding
 (and discarding the output) at least 3840&nbsp;samples (80&nbsp;ms) prior to
 the seek target in order to ensure that the output audio is correct by the
 time it reaches the seek target.
This 'pre-roll' is separate from, and unrelated to, the 'pre-skip' used at the
 beginning of the stream.
If the point 80&nbsp;ms prior to the seek target comes before the initial PCM
 sample position, an implementation SHOULD start decoding from the beginning of
 the stream, applying pre-skip as normal, regardless of whether the pre-skip is
 larger or smaller than 80&nbsp;ms, and then continue to discard samples
 to reach the seek target (if any).
</t>
</section>

</section>

<section anchor="headers" title="Header Packets">
<t>
An Ogg Opus logical stream contains exactly two mandatory header packets:
 an identification header and a comment header.
</t>

<section anchor="id_header" title="Identification Header">

<figure anchor="id_header_packet" title="ID Header Packet" align="center">
<artwork align="center"><![CDATA[
 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|      'O'      |      'p'      |      'u'      |      's'      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|      'H'      |      'e'      |      'a'      |      'd'      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|  Version = 1  | Channel Count |           Pre-skip            |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                     Input Sample Rate (Hz)                    |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|   Output Gain (Q7.8 in dB)    | Mapping Family|               |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+               :
|                                                               |
:               Optional Channel Mapping Table...               :
|                                                               |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
]]></artwork>
</figure>

<t>
The fields in the identification (ID) header have the following meaning:
<list style="numbers">
<t>Magic Signature:
<vspace blankLines="1"/>
This is an 8-octet (64-bit) field that allows codec identification and is
 human-readable.
It contains, in order, the magic numbers:
<list style="empty">
<t>0x4F 'O'</t>
<t>0x70 'p'</t>
<t>0x75 'u'</t>
<t>0x73 's'</t>
<t>0x48 'H'</t>
<t>0x65 'e'</t>
<t>0x61 'a'</t>
<t>0x64 'd'</t>
</list>
Starting with "Op" helps distinguish it from audio data packets, as this is an
 invalid TOC sequence.
<vspace blankLines="1"/>
</t>
<t>Version (8 bits, unsigned):
<vspace blankLines="1"/>
The version number MUST always be '1' for this version of the encapsulation
 specification.
Implementations SHOULD treat streams where the upper four bits of the version
 number match that of a recognized specification as backwards-compatible with
 that specification.
That is, the version number can be split into "major" and "minor" version
 sub-fields, with changes to the "minor" sub-field (in the lower four bits)
 signaling compatible changes.
For example, an implementation of this specification SHOULD accept any stream
 with a version number of '15' or less, and SHOULD assume any stream with a
 version number '16' or greater is incompatible.
The initial version '1' was chosen to keep implementations from relying on this
 octet as a null terminator for the "OpusHead" string.
<vspace blankLines="1"/>
</t>
<t>Output Channel Count 'C' (8 bits, unsigned):
<vspace blankLines="1"/>
This is the number of output channels.
This might be different than the number of encoded channels, which can change
 on a packet-by-packet basis.
This value MUST NOT be zero.
The maximum allowable value depends on the channel mapping family, and might be
 as large as 255.
See <xref target="channel_mapping"/> for details.
<vspace blankLines="1"/>
</t>
<t>Pre-skip (16 bits, unsigned, little
 endian):
<vspace blankLines="1"/>
This is the number of samples (at 48&nbsp;kHz) to discard from the decoder
 output when starting playback, and also the number to subtract from a page's
 granule position to calculate its PCM sample position.
When cropping the beginning of existing Ogg Opus streams, a pre-skip of at
 least 3,840&nbsp;samples (80&nbsp;ms) is RECOMMENDED to ensure complete
 convergence in the decoder.
<vspace blankLines="1"/>
</t>
<t>Input Sample Rate (32 bits, unsigned, little
 endian):
<vspace blankLines="1"/>
This field is <spanx style="emph">not</spanx> the sample rate to use for
 playback of the encoded data.
<vspace blankLines="1"/>
Opus can switch between internal audio bandwidths of 4, 6, 8, 12, and
 20&nbsp;kHz.
Each packet in the stream can have a different audio bandwidth.
Regardless of the audio bandwidth, the reference decoder supports decoding any
 stream at a sample rate of 8, 12, 16, 24, or 48&nbsp;kHz.
The original sample rate of the audio passed to the encoder is not preserved
 by the lossy compression.
<vspace blankLines="1"/>
An Ogg Opus player SHOULD select the playback sample rate according to the
 following procedure:
<list style="numbers">
<t>If the hardware supports 48&nbsp;kHz playback, decode at 48&nbsp;kHz.</t>
<t>Otherwise, if the hardware's highest available sample rate is a supported
 rate, decode at this sample rate.</t>
<t>Otherwise, if the hardware's highest available sample rate is less than
 48&nbsp;kHz, decode at the next highest supported rate above this and
 resample.</t>
<t>Otherwise, decode at 48&nbsp;kHz and resample.</t>
</list>
However, the 'Input Sample Rate' field allows the muxer to pass the sample
 rate of the original input stream as metadata.
This is useful when the user requires the output sample rate to match the
 input sample rate.
For example, when not playing the output, an implementation writing PCM format
 samples to disk might choose to resample the audio back to the original input
 sample rate to reduce surprise to the user, who might reasonably expect to get
 back a file with the same sample rate.
<vspace blankLines="1"/>
A value of zero indicates 'unspecified'.
Muxers SHOULD write the actual input sample rate or zero, but implementations
 which do something with this field SHOULD take care to behave sanely if given
 crazy values (e.g., do not actually upsample the output to 10 MHz if
 requested).
Implementations SHOULD support input sample rates between 8&nbsp;kHz and
 192&nbsp;kHz (inclusive).
Rates outside this range MAY be ignored by falling back to the default rate of
 48&nbsp;kHz instead.
<vspace blankLines="1"/>
</t>
<t>Output Gain (16 bits, signed, little endian):
<vspace blankLines="1"/>
This is a gain to be applied when decoding.
It is 20*log10 of the factor by which to scale the decoder output to achieve
 the desired playback volume, stored in a 16-bit, signed, two's complement
 fixed-point value with 8 fractional bits (i.e., Q7.8).
<vspace blankLines="1"/>
To apply the gain, an implementation could use
<figure align="center">
<artwork align="center"><![CDATA[
sample *= pow(10, output_gain/(20.0*256)) ,
]]></artwork>
</figure>
 where output_gain is the raw 16-bit value from the header.
<vspace blankLines="1"/>
Virtually all players and media frameworks SHOULD apply it by default.
If a player chooses to apply any volume adjustment or gain modification, such
 as the R128_TRACK_GAIN (see <xref target="comment_header"/>), the adjustment
 MUST be applied in addition to this output gain in order to achieve playback
 at the normalized volume.
<vspace blankLines="1"/>
A muxer SHOULD set this field to zero, and instead apply any gain prior to
 encoding, when this is possible and does not conflict with the user's wishes.
A nonzero output gain indicates the gain was adjusted after encoding, or that
 a user wished to adjust the gain for playback while preserving the ability
 to recover the original signal amplitude.
<vspace blankLines="1"/>
Although the output gain has enormous range (+/- 128 dB, enough to amplify
 inaudible sounds to the threshold of physical pain), most applications can
 only reasonably use a small portion of this range around zero.
The large range serves in part to ensure that gain can always be losslessly
 transferred between OpusHead and R128 gain tags (see below) without
 saturating.
<vspace blankLines="1"/>
</t>
<t>Channel Mapping Family (8 bits, unsigned):
<vspace blankLines="1"/>
This octet indicates the order and semantic meaning of the output channels.
<vspace blankLines="1"/>
Each possible value of this octet indicates a mapping family, which defines a
 set of allowed channel counts, and the ordered set of channel names for each
 allowed channel count.
The details are described in <xref target="channel_mapping"/>.
</t>
<t>Channel Mapping Table:
This table defines the mapping from encoded streams to output channels.
It MUST be omitted when the channel mapping family is 0, but is
 REQUIRED otherwise.
Its contents are specified in <xref target="channel_mapping"/>.
</t>
</list>
</t>

<t>
All fields in the ID headers are REQUIRED, except for the channel mapping
 table, which MUST be omitted when the channel mapping family is 0, but
 is REQUIRED otherwise.
Implementations SHOULD reject ID headers which do not contain enough data for
 these fields, even if they contain a valid Magic Signature.
Future versions of this specification, even backwards-compatible versions,
 might include additional fields in the ID header.
If an ID header has a compatible major version, but a larger minor version,
 an implementation MUST NOT reject it for containing additional data not
 specified here, provided it still completes on the first page.
</t>

<section anchor="channel_mapping" title="Channel Mapping">
<t>
An Ogg Opus stream allows mapping one number of Opus streams (N) to a possibly
 larger number of decoded channels (M&nbsp;+&nbsp;N) to yet another number of
 output channels (C), which might be larger or smaller than the number of
 decoded channels.
The order and meaning of these channels are defined by a channel mapping,
 which consists of the 'channel mapping family' octet and, for channel mapping
 families other than family&nbsp;0, a channel mapping table, as illustrated in
 <xref target="channel_mapping_table"/>.
</t>

<figure anchor="channel_mapping_table" title="Channel Mapping Table"
 align="center">
<artwork align="center"><![CDATA[
 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
                                                +-+-+-+-+-+-+-+-+
                                                | Stream Count  |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Coupled Count |              Channel Mapping...               :
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
]]></artwork>
</figure>

<t>
The fields in the channel mapping table have the following meaning:
<list style="numbers" counter="8">
<t>Stream Count 'N' (8 bits, unsigned):
<vspace blankLines="1"/>
This is the total number of streams encoded in each Ogg packet.
This value is necessary to correctly parse the packed Opus packets inside an
 Ogg packet, as described in <xref target="packet_organization"/>.
This value MUST NOT be zero, as without at least one Opus packet with a valid
 TOC sequence, a demuxer cannot recover the duration of an Ogg packet.
<vspace blankLines="1"/>
For channel mapping family&nbsp;0, this value defaults to 1, and is not coded.
<vspace blankLines="1"/>
</t>
<t>Coupled Stream Count 'M' (8 bits, unsigned):
This is the number of streams whose decoders are to be configured to produce
 two channels (stereo).
This MUST be no larger than the total number of streams, N.
<vspace blankLines="1"/>
Each packet in an Opus stream has an internal channel count of 1 or 2, which
 can change from packet to packet.
This is selected by the encoder depending on the bitrate and the audio being
 encoded.
The original channel count of the audio passed to the encoder is not
 necessarily preserved by the lossy compression.
<vspace blankLines="1"/>
Regardless of the internal channel count, any Opus stream can be decoded as
 mono (a single channel) or stereo (two channels) by appropriate initialization
 of the decoder.
The 'coupled stream count' field indicates that the decoders for the first M
 Opus streams are to be initialized for stereo (two-channel) output, and the
 remaining (N&nbsp;-&nbsp;M) decoders are to be initialized for mono (a single
 channel) only.
The total number of decoded channels, (M&nbsp;+&nbsp;N), MUST be no larger than
 255, as there is no way to index more channels than that in the channel
 mapping.
<vspace blankLines="1"/>
For channel mapping family&nbsp;0, this value defaults to (C&nbsp;-&nbsp;1)
 (i.e., 0 for mono and 1 for stereo), and is not coded.
<vspace blankLines="1"/>
</t>
<t>Channel Mapping (8*C bits):
This contains one octet per output channel, indicating which decoded channel
 is to be used for each one.
Let 'index' be the value of this octet for a particular output channel.
This value MUST either be smaller than (M&nbsp;+&nbsp;N), or be the special
 value 255.
If 'index' is less than 2*M, the output MUST be taken from decoding stream
 ('index'/2) as stereo and selecting the left channel if 'index' is even, and
 the right channel if 'index' is odd.
If 'index' is 2*M or larger, but less than 255, the output MUST be taken from
 decoding stream ('index'&nbsp;-&nbsp;M) as mono.
If 'index' is 255, the corresponding output channel MUST contain pure silence.
<vspace blankLines="1"/>
The number of output channels, C, is not constrained to match the number of
 decoded channels (M&nbsp;+&nbsp;N).
A single index value MAY appear multiple times, i.e., the same decoded channel
 might be mapped to multiple output channels.
Some decoded channels might not be assigned to any output channel, as well.
<vspace blankLines="1"/>
For channel mapping family&nbsp;0, the first index defaults to 0, and if
 C&nbsp;==&nbsp;2, the second index defaults to 1.
Neither index is coded.
</t>
</list>
</t>

<t>
After producing the output channels, the channel mapping family determines the
 semantic meaning of each one.
There are three defined mapping families in this specification.
</t>

<section anchor="channel_mapping_0" title="Channel Mapping Family 0">
<t>
Allowed numbers of channels: 1 or 2.
RTP mapping.
This is the same channel interpretation as <xref target="RFC7587"/>.
</t>
<t>
<list style="symbols">
<t>1 channel: monophonic (mono).</t>
<t>2 channels: stereo (left, right).</t>
</list>
Special mapping: This channel mapping value also
 indicates that the contents consists of a single Opus stream that is stereo if
 and only if C&nbsp;==&nbsp;2, with stream index&nbsp;0 mapped to output
 channel&nbsp;0 (mono, or left channel) and stream index&nbsp;1 mapped to
 output channel&nbsp;1 (right channel) if stereo.
When the 'channel mapping family' octet has this value, the channel mapping
 table MUST be omitted from the ID header packet.
</t>
</section>

<section anchor="channel_mapping_1" title="Channel Mapping Family 1">
<t>
Allowed numbers of channels: 1...8.
Vorbis channel order.
</t>
<t>
Each channel is assigned to a speaker location in a conventional surround
 arrangement.
Specific locations depend on the number of channels, and are given below
 in order of the corresponding channel indices.
<list style="symbols">
  <t>1 channel: monophonic (mono).</t>
  <t>2 channels: stereo (left, right).</t>
  <t>3 channels: linear surround (left, center, right)</t>
  <t>4 channels: quadraphonic (front&nbsp;left, front&nbsp;right, rear&nbsp;left, rear&nbsp;right).</t>
  <t>5 channels: 5.0 surround (front&nbsp;left, front&nbsp;center, front&nbsp;right, rear&nbsp;left, rear&nbsp;right).</t>
  <t>6 channels: 5.1 surround (front&nbsp;left, front&nbsp;center, front&nbsp;right, rear&nbsp;left, rear&nbsp;right, LFE).</t>
  <t>7 channels: 6.1 surround (front&nbsp;left, front&nbsp;center, front&nbsp;right, side&nbsp;left, side&nbsp;right, rear&nbsp;center, LFE).</t>
  <t>8 channels: 7.1 surround (front&nbsp;left, front&nbsp;center, front&nbsp;right, side&nbsp;left, side&nbsp;right, rear&nbsp;left, rear&nbsp;right, LFE)</t>
</list>
</t>
<t>
This set of surround options and speaker location orderings is the same
 as those used by the Vorbis codec <xref target="vorbis-mapping"/>.
The ordering is different from the one used by the
 WAVE <xref target="wave-multichannel"/> and
 FLAC <xref target="flac"/> formats,
 so correct ordering requires permutation of the output channels when decoding
 to or encoding from those formats.
'LFE' here refers to a Low Frequency Effects channel, often mapped to a
  subwoofer with no particular spatial position.
Implementations SHOULD identify 'side' or 'rear' speaker locations with
 'surround' and 'back' as appropriate when interfacing with audio formats
 or systems which prefer that terminology.
</t>
</section>

<section anchor="channel_mapping_255"
 title="Channel Mapping Family 255">
<t>
Allowed numbers of channels: 1...255.
No defined channel meaning.
</t>
<t>
Channels are unidentified.
General-purpose players SHOULD NOT attempt to play these streams.
Offline implementations MAY deinterleave the output into separate PCM files,
 one per channel.
Implementations SHOULD NOT produce output for channels mapped to stream index
 255 (pure silence) unless they have no other way to indicate the index of
 non-silent channels.
</t>
</section>

<section anchor="channel_mapping_undefined"
 title="Undefined Channel Mappings">
<t>
The remaining channel mapping families (2...254) are reserved.
An implementation encountering a reserved channel mapping family value SHOULD
 act as though the value is 255.
</t>
</section>

<section anchor="downmix" title="Downmixing">
<t>
An Ogg Opus player MUST support any valid channel mapping with a channel
 mapping family of 0 or 1, even if the number of channels does not match the
 physically connected audio hardware.
Players SHOULD perform channel mixing to increase or reduce the number of
 channels as needed.
</t>

<t>
Implementations MAY use the following matrices to implement downmixing from
 multichannel files using <xref target="channel_mapping_1">Channel Mapping
 Family 1</xref>, which are known to give acceptable results for stereo.
Matrices for 3 and 4 channels are normalized so each coefficient row sums
 to 1 to avoid clipping.
For 5 or more channels they are normalized to 2 as a compromise between
 clipping and dynamic range reduction.
</t>
<t>
In these matrices the front left and front right channels are generally
passed through directly.
When a surround channel is split between both the left and right stereo
 channels, coefficients are chosen so their squares sum to 1, which
 helps preserve the perceived intensity.
Rear channels are mixed more diffusely or attenuated to maintain focus
 on the front channels.
</t>

<figure anchor="downmix-matrix-3"
 title="Stereo downmix matrix for the linear surround channel mapping"
 align="center">
<artwork align="center"><![CDATA[
L output = ( 0.585786 * left + 0.414214 * center                    )
R output = (                   0.414214 * center + 0.585786 * right )
]]></artwork>
<postamble>
Exact coefficient values are 1 and 1/sqrt(2), multiplied by
 1/(1&nbsp;+&nbsp;1/sqrt(2)) for normalization.
</postamble>
</figure>

<figure anchor="downmix-matrix-4"
 title="Stereo downmix matrix for the quadraphonic channel mapping"
 align="center">
<artwork align="center"><![CDATA[
/          \   /                                     \ / FL \
| L output |   | 0.422650 0.000000 0.366025 0.211325 | | FR |
| R output | = | 0.000000 0.422650 0.211325 0.366025 | | RL |
\          /   \                                     / \ RR /
]]></artwork>
<postamble>
Exact coefficient values are 1, sqrt(3)/2 and 1/2, multiplied by
 1/(1&nbsp;+&nbsp;sqrt(3)/2&nbsp;+&nbsp;1/2) for normalization.
</postamble>
</figure>

<figure anchor="downmix-matrix-5"
 title="Stereo downmix matrix for the 5.0 surround mapping"
 align="center">
<artwork align="center"><![CDATA[
                                                         / FL \
/   \   /                                              \ | FC |
| L |   | 0.650802 0.460186 0.000000 0.563611 0.325401 | | FR |
| R | = | 0.000000 0.460186 0.650802 0.325401 0.563611 | | RL |
\   /   \                                              / | RR |
                                                         \    /
]]></artwork>
<postamble>
Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2 and 1/2, multiplied by
 2/(1&nbsp;+&nbsp;1/sqrt(2)&nbsp;+&nbsp;sqrt(3)/2&nbsp;+&nbsp;1/2)
 for normalization.
</postamble>
</figure>

<figure anchor="downmix-matrix-6"
 title="Stereo downmix matrix for the 5.1 surround mapping"
 align="center">
<artwork align="center"><![CDATA[
                                                                /FL \
/ \   /                                                       \ |FC |
|L|   | 0.529067 0.374107 0.000000 0.458186 0.264534 0.374107 | |FR |
|R| = | 0.000000 0.374107 0.529067 0.264534 0.458186 0.374107 | |RL |
\ /   \                                                       / |RR |
                                                                \LFE/
]]></artwork>
<postamble>
Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2 and 1/2, multiplied by
2/(1&nbsp;+&nbsp;1/sqrt(2)&nbsp;+&nbsp;sqrt(3)/2&nbsp;+&nbsp;1/2 + 1/sqrt(2))
 for normalization.
</postamble>
</figure>

<figure anchor="downmix-matrix-7"
 title="Stereo downmix matrix for the 6.1 surround mapping"
 align="center">
<artwork align="center"><![CDATA[
 /                                                                \
 | 0.455310 0.321953 0.000000 0.394310 0.227655 0.278819 0.321953 |
 | 0.000000 0.321953 0.455310 0.227655 0.394310 0.278819 0.321953 |
 \                                                                /
]]></artwork>
<postamble>
Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2, 1/2 and
 sqrt(3)/2/sqrt(2), multiplied by
 2/(1&nbsp;+&nbsp;1/sqrt(2)&nbsp;+&nbsp;sqrt(3)/2&nbsp;+&nbsp;1/2 +
 sqrt(3)/2/sqrt(2) + 1/sqrt(2)) for normalization.
The coefficients are in the same order as in <xref target="channel_mapping_1" />,
 and the matrices above.
</postamble>
</figure>

<figure anchor="downmix-matrix-8"
 title="Stereo downmix matrix for the 7.1 surround mapping"
 align="center">
<artwork align="center"><![CDATA[
/                                                                 \
| .388631 .274804 .000000 .336565 .194316 .336565 .194316 .274804 |
| .000000 .274804 .388631 .194316 .336565 .194316 .336565 .274804 |
\                                                                 /
]]></artwork>
<postamble>
Exact coefficient values are 1, 1/sqrt(2), sqrt(3)/2 and 1/2, multiplied by
 2/(2&nbsp;+&nbsp;2/sqrt(2)&nbsp;+&nbsp;sqrt(3)) for normalization.
The coefficients are in the same order as in <xref target="channel_mapping_1" />,
 and the matrices above.
</postamble>
</figure>

</section>

</section> <!-- end channel_mapping_table -->

</section> <!-- end id_header -->

<section anchor="comment_header" title="Comment Header">

<figure anchor="comment_header_packet" title="Comment Header Packet"
 align="center">
<artwork align="center"><![CDATA[
 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|      'O'      |      'p'      |      'u'      |      's'      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|      'T'      |      'a'      |      'g'      |      's'      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                     Vendor String Length                      |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                                                               |
:                        Vendor String...                       :
|                                                               |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                   User Comment List Length                    |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                 User Comment #0 String Length                 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                                                               |
:                   User Comment #0 String...                   :
|                                                               |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|                 User Comment #1 String Length                 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
:                                                               :
]]></artwork>
</figure>

<t>
The comment header consists of a 64-bit magic signature, followed by data in
 the same format as the <xref target="vorbis-comment"/> header used in Ogg
 Vorbis, except (like Ogg Theora and Speex) the final "framing bit" specified
 in the Vorbis spec is not present.
<list style="numbers">
<t>Magic Signature:
<vspace blankLines="1"/>
This is an 8-octet (64-bit) field that allows codec identification and is
 human-readable.
It contains, in order, the magic numbers:
<list style="empty">
<t>0x4F 'O'</t>
<t>0x70 'p'</t>
<t>0x75 'u'</t>
<t>0x73 's'</t>
<t>0x54 'T'</t>
<t>0x61 'a'</t>
<t>0x67 'g'</t>
<t>0x73 's'</t>
</list>
Starting with "Op" helps distinguish it from audio data packets, as this is an
 invalid TOC sequence.
<vspace blankLines="1"/>
</t>
<t>Vendor String Length (32 bits, unsigned, little endian):
<vspace blankLines="1"/>
This field gives the length of the following vendor string, in octets.
It MUST NOT indicate that the vendor string is longer than the rest of the
 packet.
<vspace blankLines="1"/>
</t>
<t>Vendor String (variable length, UTF-8 vector):
<vspace blankLines="1"/>
This is a simple human-readable tag for vendor information, encoded as a UTF-8
 string&nbsp;<xref target="RFC3629"/>.
No terminating null octet is necessary.
<vspace blankLines="1"/>
This tag is intended to identify the codec encoder and encapsulation
 implementations, for tracing differences in technical behavior.
User-facing applications can use the 'ENCODER' user comment tag to identify
 themselves.
<vspace blankLines="1"/>
</t>
<t>User Comment List Length (32 bits, unsigned, little endian):
<vspace blankLines="1"/>
This field indicates the number of user-supplied comments.
It MAY indicate there are zero user-supplied comments, in which case there are
 no additional fields in the packet.
It MUST NOT indicate that there are so many comments that the comment string
 lengths would require more data than is available in the rest of the packet.
<vspace blankLines="1"/>
</t>
<t>User Comment #i String Length (32 bits, unsigned, little endian):
<vspace blankLines="1"/>
This field gives the length of the following user comment string, in octets.
There is one for each user comment indicated by the 'user comment list length'
 field.
It MUST NOT indicate that the string is longer than the rest of the packet.
<vspace blankLines="1"/>
</t>
<t>User Comment #i String (variable length, UTF-8 vector):
<vspace blankLines="1"/>
This field contains a single user comment string.
There is one for each user comment indicated by the 'user comment list length'
 field.
</t>
</list>
</t>

<t>
The vendor string length and user comment list length are REQUIRED, and
 implementations SHOULD reject comment headers that do not contain enough data
 for these fields, or that do not contain enough data for the corresponding
 vendor string or user comments they describe.
Making this check before allocating the associated memory to contain the data
 helps prevent a possible Denial-of-Service (DoS) attack from small comment
 headers that claim to contain strings longer than the entire packet or more
 user comments than than could possibly fit in the packet.
</t>

<t>
Immediately following the user comment list, the comment header MAY
 contain zero-padding or other binary data which is not specified here.
If the least-significant bit of the first byte of this data is 1, then editors
 SHOULD preserve the contents of this data when updating the tags, but if this
 bit is 0, all such data MAY be treated as padding, and truncated or discarded
 as desired.
</t>

<t>
The comment header can be arbitrarily large and might be spread over a large
 number of Ogg pages.
Implementations SHOULD avoid attempting to allocate excessive amounts of memory
 when presented with a very large comment header.
To accomplish this, implementations MAY reject a comment header larger than
 125,829,120&nbsp;octets, and MAY ignore individual comments that are not fully
 contained within the first 61,440 octets of the comment header.
</t>

<section anchor="comment_format" title="Tag Definitions">
<t>
The user comment strings follow the NAME=value format described by
 <xref target="vorbis-comment"/> with the same recommended tag names:
 ARTIST, TITLE, DATE, ALBUM, and so on.
</t>
<t>
Two new comment tags are introduced here:
</t>

<t>First, an optional gain for track normalization:</t>
<figure align="center">
<artwork align="left"><![CDATA[
R128_TRACK_GAIN=-573
]]></artwork>
</figure>
<t>
 representing the volume shift needed to normalize the track's volume
 during isolated playback, in random shuffle, and so on.
The gain is a Q7.8 fixed point number in dB, as in the ID header's 'output
 gain' field.
This tag is similar to the REPLAYGAIN_TRACK_GAIN tag in
 Vorbis&nbsp;<xref target="replay-gain"/>, except that the normal volume
 reference is the <xref target="EBU-R128"/> standard.
</t>
<t>Second, an optional gain for album normalization:</t>
<figure align="center">
<artwork align="left"><![CDATA[
R128_ALBUM_GAIN=111
]]></artwork>
</figure>
<t>
 representing the volume shift needed to normalize the overall volume when
 played as part of a particular collection of tracks.
The gain is also a Q7.8 fixed point number in dB, as in the ID header's
 'output gain' field.
</t>
<t>
An Ogg Opus stream MUST NOT have more than one of each tag, and if present
 their values MUST be an integer from -32768 to 32767, inclusive,
 represented in ASCII as a base 10 number with no whitespace.
A leading '+' or '-' character is valid.
Leading zeros are also permitted, but the value MUST be represented by
 no more than 6 characters.
Other non-digit characters MUST NOT be present.
</t>
<t>
If present, R128_TRACK_GAIN and R128_ALBUM_GAIN MUST correctly represent
 the R128 normalization gain relative to the 'output gain' field specified
 in the ID header.
If a player chooses to make use of the R128_TRACK_GAIN tag or the
 R128_ALBUM_GAIN tag, it MUST apply those gains
 <spanx style="emph">in addition</spanx> to the 'output gain' value.
If a tool modifies the ID header's 'output gain' field, it MUST also update or
 remove the R128_TRACK_GAIN and R128_ALBUM_GAIN comment tags if present.
A muxer SHOULD assume that by default tools will respect the 'output gain'
 field, and not the comment tag.
</t>
<t>
To avoid confusion with multiple normalization schemes, an Opus comment header
 SHOULD NOT contain any of the REPLAYGAIN_TRACK_GAIN, REPLAYGAIN_TRACK_PEAK,
 REPLAYGAIN_ALBUM_GAIN, or REPLAYGAIN_ALBUM_PEAK tags.
<xref target="EBU-R128"/> normalization is preferred to the earlier
 REPLAYGAIN schemes because of its clear definition and adoption by industry.
Peak normalizations are difficult to calculate reliably for lossy codecs
 because of variation in excursion heights due to decoder differences.
In the authors' investigations they were not applied consistently or broadly
 enough to merit inclusion here.
</t>
</section> <!-- end comment_format -->
</section> <!-- end comment_header -->

</section> <!-- end headers -->

<section anchor="packet_size_limits" title="Packet Size Limits">
<t>
Technically, valid Opus packets can be arbitrarily large due to the padding
 format, although the amount of non-padding data they can contain is bounded.
These packets might be spread over a similarly enormous number of Ogg pages.
When encoding, implementations SHOULD limit the use of padding in audio data
 packets to no more than is necessary to make a variable bitrate (VBR) stream
 constant bitrate (CBR).
Demuxers SHOULD reject audio data packets larger than 61,440 octets per
 Opus stream.
Such packets necessarily contain more padding than needed for this purpose.
Demuxers SHOULD avoid attempting to allocate excessive amounts of memory when
 presented with a very large packet.
Demuxers MAY reject or partially process audio data packets larger than
 61,440&nbsp;octets in an Ogg Opus stream with channel mapping families&nbsp;0
 or&nbsp;1.
Demuxers MAY reject or partially process audio data packets in any Ogg Opus
 stream if the packet is larger than 61,440&nbsp;octets and also larger than
 7,680&nbsp;octets per Opus stream.
The presence of an extremely large packet in the stream could indicate a
 memory exhaustion attack or stream corruption.
</t>
<t>
In an Ogg Opus stream, the largest possible valid packet that does not use
 padding has a size of (61,298*N&nbsp;-&nbsp;2) octets.
With 255&nbsp;streams, this is 15,630,988&nbsp;octets and can
 span up to 61,298&nbsp;Ogg pages, all but one of which will have a granule
 position of -1.
This is of course a very extreme packet, consisting of 255&nbsp;streams, each
 containing 120&nbsp;ms of audio encoded as 2.5&nbsp;ms frames, each frame
 using the maximum possible number of octets (1275) and stored in the least
 efficient manner allowed (a VBR code&nbsp;3 Opus packet).
Even in such a packet, most of the data will be zeros as 2.5&nbsp;ms frames
 cannot actually use all 1275&nbsp;octets.
</t>
<t>
The largest packet consisting of entirely useful data is
 (15,326*N&nbsp;-&nbsp;2) octets.
This corresponds to 120&nbsp;ms of audio encoded as 10&nbsp;ms frames in either
 SILK or Hybrid mode, but at a data rate of over 1&nbsp;Mbps, which makes little
 sense for the quality achieved.
</t>
<t>
A more reasonable limit is (7,664*N&nbsp;-&nbsp;2) octets.
This corresponds to 120&nbsp;ms of audio encoded as 20&nbsp;ms stereo CELT mode
 frames, with a total bitrate just under 511&nbsp;kbps (not counting the Ogg
 encapsulation overhead).
For channel mapping family 1, N=8 provides a reasonable upper bound, as it
 allows for each of the 8 possible output channels to be decoded from a
 separate stereo Opus stream.
This gives a size of 61,310&nbsp;octets, which is rounded up to a multiple of
 1,024&nbsp;octets to yield the audio data packet size of 61,440&nbsp;octets
 that any implementation is expected to be able to process successfully.
</t>
</section>

<section anchor="encoder" title="Encoder Guidelines">
<t>
When encoding Opus streams, Ogg muxers SHOULD take into account the
 algorithmic delay of the Opus encoder.
</t>
<t>
In encoders derived from the reference implementation, the number of
 samples can be queried with:
</t>
<figure align="center">
<artwork align="center"><![CDATA[
 opus_encoder_ctl(encoder_state, OPUS_GET_LOOKAHEAD(&delay_samples));
]]></artwork>
</figure>
<t>
To achieve good quality in the very first samples of a stream, implementations
 MAY use linear predictive coding (LPC) extrapolation
 <xref target="linear-prediction"/> to generate at least 120 extra samples at
 the beginning to avoid the Opus encoder having to encode a discontinuous
 signal.
For an input file containing 'length' samples, the implementation SHOULD set
 the pre-skip header value to (delay_samples&nbsp;+&nbsp;extra_samples), encode
 at least (length&nbsp;+&nbsp;delay_samples&nbsp;+&nbsp;extra_samples)
 samples, and set the granule position of the last page to
 (length&nbsp;+&nbsp;delay_samples&nbsp;+&nbsp;extra_samples).
This ensures that the encoded file has the same duration as the original, with
 no time offset. The best way to pad the end of the stream is to also use LPC
 extrapolation, but zero-padding is also acceptable.
</t>

<section anchor="lpc" title="LPC Extrapolation">
<t>
The first step in LPC extrapolation is to compute linear prediction
 coefficients. <xref target="lpc-sample"/>
When extending the end of the signal, order-N (typically with N ranging from 8
 to 40) LPC analysis is performed on a window near the end of the signal.
The last N samples are used as memory to an infinite impulse response (IIR)
 filter.
</t>
<t>
The filter is then applied on a zero input to extrapolate the end of the signal.
Let a(k) be the kth LPC coefficient and x(n) be the nth sample of the signal,
 each new sample past the end of the signal is computed as:
</t>
<figure align="center">
<artwork align="center"><![CDATA[
        N
       ---
x(n) = \   a(k)*x(n-k)
       /
       ---
       k=1
]]></artwork>
</figure>
<t>
The process is repeated independently for each channel.
It is possible to extend the beginning of the signal by applying the same
 process backward in time.
When extending the beginning of the signal, it is best to apply a "fade in" to
 the extrapolated signal, e.g. by multiplying it by a half-Hanning window
 <xref target="hanning"/>.
</t>

</section>

<section anchor="continuous_chaining" title="Continuous Chaining">
<t>
In some applications, such as Internet radio, it is desirable to cut a long
 stream into smaller chains, e.g. so the comment header can be updated.
This can be done simply by separating the input streams into segments and
 encoding each segment independently.
The drawback of this approach is that it creates a small discontinuity
 at the boundary due to the lossy nature of Opus.
A muxer MAY avoid this discontinuity by using the following procedure:
<list style="numbers">
<t>Encode the last frame of the first segment as an independent frame by
 turning off all forms of inter-frame prediction.
De-emphasis is allowed.</t>
<t>Set the granule position of the last page to a point near the end of the
 last frame.</t>
<t>Begin the second segment with a copy of the last frame of the first
 segment.</t>
<t>Set the pre-skip value of the second stream in such a way as to properly
 join the two streams.</t>
<t>Continue the encoding process normally from there, without any reset to
 the encoder.</t>
</list>
</t>
<t>
In encoders derived from the reference implementation, inter-frame prediction
 can be turned off by calling:
</t>
<figure align="center">
<artwork align="center"><![CDATA[
 opus_encoder_ctl(encoder_state, OPUS_SET_PREDICTION_DISABLED(1));
]]></artwork>
</figure>
<t>
For best results, this implementation requires that prediction be explicitly
 enabled again before resuming normal encoding, even after a reset.
</t>

</section>

</section>

<section anchor="implementation" title="Implementation Status">
<t>
A brief summary of major implementations of this draft is available
 at <eref target="https://wiki.xiph.org/OggOpusImplementation"/>,
 along with their status.
</t>
<t>
[Note to RFC Editor: please remove this entire section before
 final publication per <xref target="RFC6982"/>, along with
 its references.]
</t>
</section>

<section anchor="security" title="Security Considerations">
<t>
Implementations of the Opus codec need to take appropriate security
 considerations into account, as outlined in <xref target="RFC4732"/>.
This is just as much a problem for the container as it is for the codec itself.
Robustness against malicious payloads is extremely important.
Malicious payloads MUST NOT cause an implementation to overrun its allocated
 memory or to take an excessive amount of resources to decode.
Although problems in encoding applications are typically rarer, the same
 applies to the muxer.
Malicious audio input streams MUST NOT cause an implementation to overrun its
 allocated memory or consume excessive resources because this would allow an
 attacker to attack transcoding gateways.
</t>

<t>
Like most other container formats, Ogg Opus streams SHOULD NOT be used with
 insecure ciphers or cipher modes that are vulnerable to known-plaintext
 attacks.
Elements such as the Ogg page capture pattern and the magic signatures in the
 ID header and the comment header all have easily predictable values, in
 addition to various elements of the codec data itself.
</t>
</section>

<section anchor="content_type" title="Content Type">
<t>
An "Ogg Opus file" consists of one or more sequentially multiplexed segments,
 each containing exactly one Ogg Opus stream.
The RECOMMENDED mime-type for Ogg Opus files is "audio/ogg".
</t>

<t>
If more specificity is desired, one MAY indicate the presence of Opus streams
 using the codecs parameter defined in <xref target="RFC6381"/> and
 <xref target="RFC5334"/>, e.g.,
</t>
<figure>
<artwork align="center"><![CDATA[
    audio/ogg; codecs=opus
]]></artwork>
</figure>
<t>
 for an Ogg Opus file.
</t>

<t>
The RECOMMENDED filename extension for Ogg Opus files is '.opus'.
</t>

<t>
When Opus is concurrently multiplexed with other streams in an Ogg container,
 one SHOULD use one of the "audio/ogg", "video/ogg", or "application/ogg"
 mime-types, as defined in <xref target="RFC5334"/>.
Such streams are not strictly "Ogg Opus files" as described above,
 since they contain more than a single Opus stream per sequentially
 multiplexed segment, e.g. video or multiple audio tracks.
In such cases the the '.opus' filename extension is NOT RECOMMENDED.
</t>

<t>
In either case, this document updates <xref target="RFC5334"/>
 to add 'opus' as a codecs parameter value with char[8]: 'OpusHead'
 as Codec Identifier.
</t>
</section>

<section title="IANA Considerations">
<t>
This document updates the IANA Media Types registry to add .opus
 as a file extension for "audio/ogg", and to add itself as a reference
 alongside <xref target="RFC5334"/> for "audio/ogg", "video/ogg", and
 "application/ogg" Media Types.
</t>
</section>

<section anchor="Acknowledgments" title="Acknowledgments">
<t>
Thanks to Mark Harris, Greg Maxwell, Christopher "Monty" Montgomery, and
 Jean-Marc Valin for their valuable contributions to this document.
Additional thanks to Andrew D'Addesio, Greg Maxwell, and Vincent Penquerc'h for
 their feedback based on early implementations.
</t>
</section>

<section title="Copying Conditions">
<t>
The authors agree to grant third parties the irrevocable right to copy, use,
 and distribute the work, with or without modification, in any medium, without
 royalty, provided that, unless separate permission is granted, redistributed
 modified works do not contain misleading author, version, name of work, or
 endorsement information.
</t>
</section>

</middle>
<back>
<references title="Normative References">
 &rfc2119;
 &rfc3533;
 &rfc3629;
 &rfc5334;
 &rfc6381;
 &rfc6716;

<reference anchor="EBU-R128" target="https://tech.ebu.ch/loudness">
<front>
  <title>Loudness Recommendation EBU R128</title>
  <author>
    <organization>EBU Technical Committee</organization>
  </author>
  <date month="August" year="2011"/>
</front>
</reference>

<reference anchor="vorbis-comment"
 target="https://www.xiph.org/vorbis/doc/v-comment.html">
<front>
<title>Ogg Vorbis I Format Specification: Comment Field and Header
 Specification</title>
<author initials="C." surname="Montgomery"
 fullname="Christopher &quot;Monty&quot; Montgomery"/>
<date month="July" year="2002"/>
</front>
</reference>

</references>

<references title="Informative References">

<!--?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.3550.xml"?-->
 &rfc4732;
 &rfc6982;
 &rfc7587;

<reference anchor="flac"
 target="https://xiph.org/flac/format.html">
  <front>
    <title>FLAC - Free Lossless Audio Codec Format Description</title>
    <author initials="J." surname="Coalson" fullname="Josh Coalson"/>
    <date month="January" year="2008"/>
  </front>
</reference>

<reference anchor="hanning"
 target="https://en.wikipedia.org/wiki/Hamming_function#Hann_.28Hanning.29_window">
  <front>
    <title>Hann window</title>
    <author>
      <organization>Wikipedia</organization>
    </author>
    <date month="May" year="2013"/>
  </front>
</reference>

<reference anchor="linear-prediction"
 target="https://en.wikipedia.org/wiki/Linear_predictive_coding">
  <front>
    <title>Linear Predictive Coding</title>
    <author>
      <organization>Wikipedia</organization>
    </author>
    <date month="January" year="2014"/>
  </front>
</reference>

<reference anchor="lpc-sample"
  target="https://svn.xiph.org/trunk/vorbis/lib/lpc.c">
<front>
  <title>Autocorrelation LPC coeff generation algorithm
    (Vorbis source code)</title>
<author initials="J." surname="Degener" fullname="Jutta Degener"/>
<author initials="C." surname="Bormann" fullname="Carsten Bormann"/>
<date month="November" year="1994"/>
</front>
</reference>


<reference anchor="replay-gain"
 target="https://wiki.xiph.org/VorbisComment#Replay_Gain">
<front>
<title>VorbisComment: Replay Gain</title>
<author initials="C." surname="Parker" fullname="Conrad Parker"/>
<author initials="M." surname="Leese" fullname="Martin Leese"/>
<date month="June" year="2009"/>
</front>
</reference>

<reference anchor="seeking"
 target="https://wiki.xiph.org/Seeking">
<front>
<title>Granulepos Encoding and How Seeking Really Works</title>
<author initials="S." surname="Pfeiffer" fullname="Silvia Pfeiffer"/>
<author initials="C." surname="Parker" fullname="Conrad Parker"/>
<author initials="G." surname="Maxwell" fullname="Greg Maxwell"/>
<date month="May" year="2012"/>
</front>
</reference>

<reference anchor="vorbis-mapping"
 target="https://www.xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-810004.3.9">
<front>
<title>The Vorbis I Specification, Section 4.3.9 Output Channel Order</title>
<author initials="C." surname="Montgomery"
 fullname="Christopher &quot;Monty&quot; Montgomery"/>
<date month="January" year="2010"/>
</front>
</reference>

<reference anchor="vorbis-trim"
 target="https://xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-132000A.2">
  <front>
    <title>The Vorbis I Specification, Appendix&nbsp;A: Embedding Vorbis
      into an Ogg stream</title>
    <author initials="C." surname="Montgomery"
     fullname="Christopher &quot;Monty&quot; Montgomery"/>
    <date month="November" year="2008"/>
  </front>
</reference>

<reference anchor="wave-multichannel"
 target="http://msdn.microsoft.com/en-us/windows/hardware/gg463006.aspx">
  <front>
    <title>Multiple Channel Audio Data and WAVE Files</title>
    <author>
      <organization>Microsoft Corporation</organization>
    </author>
    <date month="March" year="2007"/>
  </front>
</reference>

</references>

</back>
</rfc>