shaka-packager/packager/mpd/base/media_info.proto

// This file defines the protocol between Muxer and classes for generating MPD
// files.

syntax = "proto2";

package shaka;

message Range {
  optional uint64 begin = 1;
  optional uint64 end = 2;
}

message MediaInfo {
  enum ContainerType {
    CONTAINER_UNKNOWN = 0;
    CONTAINER_MP4 = 1;
    CONTAINER_MPEG2_TS= 2;
    CONTAINER_WEBM = 3;
    CONTAINER_TEXT = 4;
    CONTAINER_PACKED_AUDIO = 5;
  }

  message VideoInfo {
    optional string codec = 1;

    // The width and height of the actual number of pixels. This will not be the
    // same as the visual width and height if the sample aspect ratio (sar)
    // is not 1:1.
    optional uint32 width = 2;
    optional uint32 height = 3;

    optional uint32 time_scale = 4;
    // Relative to |time_scale|. IOW |time_scale| / |frame_duration| is the
    // framerate.
    optional uint64 frame_duration = 5;
    optional bytes decoder_config = 6;

    // pixel_width:pixel_height is the the sample aspect ratio (sar) of the
    // video.
    // Note that (pixel_width * width):(pixel_height * height) is the picture
    // aspect ratio, or the @par attribute set on AdaptationSet element.
    optional uint32 pixel_width = 7;
    optional uint32 pixel_height = 8;

    // playback_rate: the playout capability (e.g., 4x, 8x, 16x fast foward) of
    // the trick play stream.
    optional uint32 playback_rate = 9;

    // Transfer characteristics. Useful to determine the VIDEO-RANGE for HLS,
    // i.e. whether it is SDR or HDR.
    optional uint32 transfer_characteristics = 10;
  }

  message AudioInfo {
    optional string codec = 1;
    optional uint32 sampling_frequency = 2;
    optional uint32 time_scale = 3;
    optional uint32 num_channels = 4;
    optional string language = 5;
    optional bytes decoder_config = 6;
    optional AudioCodecSpecificData codec_specific_data = 7;
  }

  message AudioCodecSpecificData {
    // EC3 Channel map bit fields, encoded based on ETSI TS 102 366 V1.3.1
    // Digital Audio Compression (AC-3, Enhanced AC-3) Standard E.1.3.1.8.
    // Or AC4 Channel mask bit fields, encoded based on ETSI TS 103 190-2
    // V1.2.1 Digital Audio Compression (AC-4) Standard; Part 2: Immersive and
    // personalized audio E.10.14.
    optional uint32 channel_mask = 1;

    // EC3 Channel configuration descriptor with MPEG scheme fields,
    // encoded based on ETSI TS 102 366 V1.4.1 Digital Audio Compression
    // (AC-3, Enhanced AC-3) Standard I.1.2.1.
    // Or AC4 Channel configuration descriptor with MPEG scheme fields,
    // encoded based on ETSI TS 103 190-2 V1.2.1 Digital Audio Compression
    // (AC-4) Standard; Part 2: Immersive and personalized audio G.3.2.
    optional uint32 channel_mpeg_value = 2;

    // Dolby Digital Plus JOC decoding complexity fields, ETSI TS 103 420 v1.2.1
    // Backwards-compatible object audio carriage using Enhanced AC-3 Standard
    // C.3.2.3.
    optional uint32 ec3_joc_complexity = 3;

    // AC4 Immersive stereo flag field, based on Dolby AC-4 in MPEG-DASH for
    // Online Delivery Specification 2.5.3.
    // https://developer.dolby.com/tools-media/online-delivery-kits/dolby-ac-4/
    optional bool ac4_ims_flag = 4;

    // AC4 Channel-based audio (CBI) flag field, encoded based on
    // ETSI TS 103 190-2 Digital Audio Compression (AC-4) Standard;
    // Part 2: Immersive and personalized audio 4.3.
    optional bool ac4_cbi_flag = 5;
  }

  message TextInfo {
    enum TextType {
      UNKNOWN = 0;
      CAPTION = 1;
      SUBTITLE = 2;
    }
    optional string codec = 1;
    optional string language = 2;
    optional TextType type = 3;
  }

  message ProtectedContent {
    message ContentProtectionEntry {
      // Human readable UUID of the DRM.
      optional string uuid = 1;
      // Human readable DRM name and version string.
      // e.g. "My Content Protection v1.0"
      optional string name_version = 2;
      // The raw 'pssh' box for the media.
      optional bytes pssh = 3;
    }

    // The default key ID for the encrypted media.
    optional bytes default_key_id = 1;
    repeated ContentProtectionEntry content_protection_entry = 2;
    // Specifies the protection scheme: 'cenc', 'cens', 'cbc1', 'cbcs'.
    // "cbca" is also valid which is a place holder for SAMPLE-AES encryption.
    optional string protection_scheme = 3 [default = 'cenc'];
    optional bool include_mspr_pro = 4 [default = true];
  }

  // TODO(rkuroiwa): Remove this. <ContentProtection> element that must be added
  // should be done by directly using the MpdBuilder interface.
  // Use this to specify ContentProtection elements that should be set in
  // the MPD, if ContentProtectionEntry is not sufficient.
  message ContentProtectionXml {
    message AttributeNameValuePair {
      optional string name = 1;
      optional string value = 2;
    }

    message Element {
      optional string name = 1;
      repeated AttributeNameValuePair attributes = 2;
      repeated Element subelements = 3;
    }

    // These two string fields are specified by the MPD spec. Just for
    // clarification, the string set in |value| is the rhs of 'value' field in
    // <ContentProtection>.
    optional string scheme_id_uri = 1;
    optional string value = 2;
    repeated AttributeNameValuePair attributes = 3;
    repeated Element subelements = 4;
  }

  optional uint32 bandwidth = 1;

  // Note that DASH IOP v3.0 explicitly mentions that a segment should only
  // have one {video, audio, text} track.
  optional VideoInfo video_info = 2;
  optional AudioInfo audio_info = 3;
  optional TextInfo text_info = 4;
  repeated ContentProtectionXml content_protections = 5;

  // This is set if the content is protected with a content protection,
  // i.e. encrypted.
  optional ProtectedContent protected_content = 15;

  // This is the reference time scale if there are multiple VideoInfo and/or
  // AudioInfo.
  optional uint32 reference_time_scale = 13;
  optional uint64 presentation_time_offset = 16;
  optional ContainerType container_type = 14 [default = CONTAINER_UNKNOWN];

  // VOD only.
  optional Range init_range = 6;
  optional Range index_range = 7;
  optional string media_file_name = 8;
  repeated Range subsegment_ranges = 23;
  // END VOD only.

  // VOD and static LIVE.
  optional float media_duration_seconds = 9;

  // LIVE only.
  optional string init_segment_name = 10;
  optional string segment_template = 11;
  // This value is the user input “segment duration”.
  // This value is not necessarily the same as the value passed to
  // MpdNotifier::NotifyNewSegment().
  optional float segment_duration_seconds = 12 [deprecated = true];
  // END LIVE only.

  // URL fields for the corresponding file_name fields above.
  // The file names are adjusted to be relative to DASH MPD or HLS media
  // playlist, or with base url prepended.
  optional string media_file_url = 17;
  optional string init_segment_url = 18;
  optional string segment_template_url = 19;

  // HLS only. Defines CHARACTERISTICS attribute of the stream.
  repeated string hls_characteristics = 20;

  // DASH only. Defines Accessibility elements of the stream. It should be in
  // the format: scheme_id_uri=value.
  repeated string dash_accessibilities = 21;
  // DASH only. Defines Role elements of the stream. The value can be a valid
  // Role value defined in "urn:mpeg:dash:role:2011" scheme or in the format:
  // scheme_id_uri=value (to be implemented).
  repeated string dash_roles = 22;

  // LOW LATENCY DASH only. Defines the availabilityTimeOffset in seconds.
  // Equal to the segment time minus the chunk duration.
  optional double availability_time_offset = 24;
  // LOW LATENCY DASH only. Defines the segment duration
  // with respect to the reference time scale.
  // Equal to the target segment duration times the reference time scale.
  optional uint64 segment_duration = 25;
}