.ISM File Format – Server manifest elements

Smooth Streaming server manifest

The server manifest is a small SMIL 2.0 document in the namespace http://www.w3.org/2001/SMIL20/Language. Microsoft documented the on-demand form for IIS Media Services, in pages now archived, and specifies the live form in MS-SSTR, where an encoder sends it to the server. An on-demand presentation is an .ism file beside its .ismv and .isma media and its .ismc client manifest; a live one is an .isml publishing point.

Elements

ElementInContentMeaning
smilroothead, bodyThe presentation.
headsmilmeta elementsPresentation-level metadata.
metaheadempty; name and content, both requiredOne item of metadata (below).
bodysmilswitchHolds the tracks.
switchbodyvideo, audio, textstreamThe tracks a player can choose between.
video, audio, textstreamswitchone or more paramOne track of that kind.
parama trackempty; name, value and valuetype="data"One track parameter (below).

meta names

namecontent
titleThe title of the presentation.
clientManifestRelativePathThe client manifest’s path relative to the server manifest, such as film.ismc. On-demand only; a live manifest leaves it out. Microsoft’s archived reference spells it clientManfiestPath, but files and Microsoft’s own published examples use clientManifestRelativePath.

Track attributes

AttributeMeaning
srcThe media file holding the track, such as film_3000.ismv. Required on demand. In a live manifest it should be Streams, the same for every track.
systemBitrateThe track’s bit rate in bits per second, required. It is the number in the QualityLevels() part of fragment addresses.
systemLanguageThe track’s language, as an RFC 1766 tag such as en. Informative only: tracks at the same bit rate in different languages still need different track names.

Track parameters

namevalue
trackIDRequired. The track’s track_ID in the trak and tfhd boxes of the media file.
trackNameThe name players use in Fragments(name=…). Tracks with the same bit rate and attributes need different names. Default video, audio or textstream.
timescaleUnits per second of the track’s times. Default 10000000 (100 nanoseconds).
manifestOutputWhether the track’s sample data is gathered into the client manifest, so it reaches players at once: TRUE/FALSE in MS-SSTR, 1/0 in the archived IIS pages. Default false.
parentTrackNameMakes this a sparse track attached to the named track, which must be neither sparse nor a control track.
SubtypeInformation a player can use to identify characteristics of the track (MS-SSTR, live).
hardwareProfileThe decoding capability a stream needs, when several streams share a bit rate: (width × height ÷ (4096 × 2048) × 100) × 1000 plus a codec complexity from 1 to 999, each division rounded down; at most 100000. Archived IIS pages only.
namespace_attributeA custom attribute that tells apart tracks with the same name and bit rate, as CustomAttributes do in the client manifest. A namespace of your own avoids clashes with future names.

Worked example

<?xml version="1.0" encoding="utf-8"?>
<smil xmlns="http://www.w3.org/2001/SMIL20/Language">
  <head>
    <meta name="clientManifestRelativePath" content="film.ismc"/>
  </head>
  <body>
    <switch>
      <video src="film_3000.ismv" systemBitrate="3000000">
        <param name="trackID" value="2" valuetype="data"/>
      </video>
      <video src="film_1000.ismv" systemBitrate="1000000">
        <param name="trackID" value="2" valuetype="data"/>
      </video>
      <audio src="film_3000.ismv" systemBitrate="128000" systemLanguage="en">
        <param name="trackID" value="1" valuetype="data"/>
        <param name="trackName" value="audio_en" valuetype="data"/>
      </audio>
    </switch>
  </body>
</smil>
Two video bit rates in two files, and English audio stored in the first file as track 1. A player asks for film.ism/QualityLevels(128000)/Fragments(audio_en=0) to get the first audio fragment.