V1.1 test vectors

proto/audio_element.proto

@@ -50,8 +50,28 @@ enum LoudspeakerLayout {
   LOUDSPEAKER_LAYOUT_7_1_4_CH = 8;
   LOUDSPEAKER_LAYOUT_3_1_2_CH = 9;
   LOUDSPEAKER_LAYOUT_BINAURAL = 10;
-  LOUDSPEAKER_LAYOUT_RESERVED_BEGIN = 11;
-  LOUDSPEAKER_LAYOUT_RESERVED_END = 16;
+  LOUDSPEAKER_LAYOUT_RESERVED_10 = 11;
+  LOUDSPEAKER_LAYOUT_RESERVED_14 = 15;
+  LOUDSPEAKER_LAYOUT_EXPANDED = 16;
+}
+
+enum ExpandedLoudspeakerLayout {
+  EXPANDED_LOUDSPEAKER_LAYOUT_INVALID = 0;
+  EXPANDED_LOUDSPEAKER_LAYOUT_LFE = 1;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_S = 2;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_SS = 3;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_RS = 4;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_TF = 5;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_TB = 6;
+  EXPANDED_LOUDSPEAKER_LAYOUT_TOP_4_CH = 7;
+  EXPANDED_LOUDSPEAKER_LAYOUT_3_0_CH = 8;
+  EXPANDED_LOUDSPEAKER_LAYOUT_9_1_6_CH = 9;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_F = 10;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_SI = 11;
+  EXPANDED_LOUDSPEAKER_LAYOUT_STEREO_TP_SI = 12;
+  EXPANDED_LOUDSPEAKER_LAYOUT_TOP_6_CH = 13;
+  EXPANDED_LOUDSPEAKER_LAYOUT_RESERVED_13 = 14;
+  EXPANDED_LOUDSPEAKER_LAYOUT_RESERVED_255 = 256;
 }
 
 message ChannelAudioLayerConfig {
@@ -65,6 +85,7 @@ message ChannelAudioLayerConfig {
   optional uint32 output_gain_flag = 7;
   optional uint32 reserved_b = 8;
   optional int32 output_gain = 9;
+  optional ExpandedLoudspeakerLayout expanded_loudspeaker_layout = 11;
 }
 
 message ScalableChannelLayoutConfig {

proto/codec_config.proto

@@ -37,10 +37,20 @@ enum OpusApplicationFlag {
 
 // Settings to configure `libopus`.
 message OpusEncoderMetadata {
-  optional int32 target_bitrate_per_channel = 1;
   optional OpusApplicationFlag application = 2;
   optional bool use_float_api = 3 [default = true];
+
+  // Fields to control the bitrate.
+  //
+  //  - If the substream is present in `substream_id_to_bitrate_override`, the
+  //  bitrate override is used.
+  //  - Otherwise a bitrate is calculated based on the number of channels.
+  //    - One channel: `target_bitrate_per_channel`.
+  //    - Two channels: `target_bitrate_per_channel * 2 *
+  //  coupling_rate_adjustment`.
+  optional int32 target_bitrate_per_channel = 1;
   optional float coupling_rate_adjustment = 4 [default = 1.0];
+  map<uint32, int32> substream_id_to_bitrate_override = 5;
 }
 
 message OpusDecoderConfig {

proto/ia_sequence_header.proto

@@ -19,6 +19,7 @@ enum ProfileVersion {
   PROFILE_VERSION_SIMPLE = 1;
   PROFILE_VERSION_BASE = 2;
   PROFILE_VERSION_BASE_ENHANCED = 3;
+  PROFILE_VERSION_RESERVED_255 = 256;
 }
 
 message IASequenceHeaderObuMetadata {

proto/mix_presentation.proto

@@ -16,14 +16,20 @@ import "obu_header.proto";
 import "param_definitions.proto";
 
 message MixPresentationAnnotations {
+  option deprecated = true;
+
   optional string mix_presentation_friendly_label = 1;
 }
 
 message MixPresentationElementAnnotations {
+  option deprecated = true;
+
   optional string audio_element_friendly_label = 1;
 }
 
 message ElementMixConfig {
+  option deprecated = true;
+
   optional MixGainParamDefinition mix_gain = 1;
 }
 
@@ -45,13 +51,21 @@ message RenderingConfig {
 message SubMixAudioElement {
   optional uint32 audio_element_id = 1;
   reserved 2;
+  // Superseded by `localized_element_annotations`.
   repeated MixPresentationElementAnnotations
-      mix_presentation_element_annotations_array = 4;
+      mix_presentation_element_annotations_array = 4 [deprecated = true];
+  // Length should be equal to `count_label` in the OBU.
+  repeated string localized_element_annotations = 6;
+
   optional RenderingConfig rendering_config = 5;
-  optional ElementMixConfig element_mix_config = 3;
+  // Superseded by `element_mix_gain`.
+  optional ElementMixConfig element_mix_config = 3 [deprecated = true];
+  optional MixGainParamDefinition element_mix_gain = 7;
 }
 
 message OutputMixConfig {
+  option deprecated = true;
+
   optional MixGainParamDefinition output_mix_gain = 1;
 }
 
@@ -147,18 +161,47 @@ message MixPresentationLayout {
 message MixPresentationSubMix {
   optional uint32 num_audio_elements = 1;
   repeated SubMixAudioElement audio_elements = 2;
-  optional OutputMixConfig output_mix_config = 3;
+
+  // Superseded by `output_mix_gain`.
+  optional OutputMixConfig output_mix_config = 3 [deprecated = true];
+  optional MixGainParamDefinition output_mix_gain = 6;
+
   optional uint32 num_layouts = 4;
   repeated MixPresentationLayout layouts = 5;
 }
 
+message MixPresentationTag {
+  optional string tag_name = 1;
+  optional string tag_value = 2;
+}
+
+message MixPresentationTags {
+  optional uint32 num_tags = 1;
+  repeated MixPresentationTag tags = 2;
+}
+
 message MixPresentationObuMetadata {
+  reserved 2;
   optional uint32 mix_presentation_id = 1;
   optional uint32 count_label = 5;
-  repeated string language_labels = 6;
-  reserved 2;
-  repeated MixPresentationAnnotations mix_presentation_annotations_array = 7;
+
+  // Superseded by `annotations_language`.
+  repeated string language_labels = 6 [deprecated = true];
+  // Length should be equal to `count_label`.
+  repeated string annotations_language = 9;
+
+  // Superseded by `localized_presentation_annotations`.
+  repeated MixPresentationAnnotations mix_presentation_annotations_array = 7
+      [deprecated = true];
+  // Length should be equal to `count_label`.
+  repeated string localized_presentation_annotations = 10;
+
   optional uint32 num_sub_mixes = 3;
   repeated MixPresentationSubMix sub_mixes = 4;
+
+  // When true, `mix_presentation_tags` will be examined.
+  optional bool include_mix_presentation_tags = 11;
+  optional MixPresentationTags mix_presentation_tags = 12;
+
   optional ObuHeaderMetadata obu_header = 8;
 }

tests/README.md

@@ -63,6 +63,33 @@ Title                                                | Summary
 `sine_1000_48kHz_512ms.wav`                          | Sine wave.                                                                                                          | 2        | 48kHz       | pcm_s16le | 512ms
 `sine_1000_48kHz.wav`                                | Sine wave.                                                                                                          | 2        | 48kHz       | pcm_s16le | 500ms
 `sine_1000_4oa_48kHz.wav`                            | Sine wave using fourth-order ambisonics.                                                                            | 25       | 48kHz       | pcm_s16le | 5000ms
+`sine_1500_stereo_48khz_-15dBFS.wav`                 | Sine wave using at -15dBFS.                                                                                         | 2        | 48kHz       | pcm_s16le | 5000ms
 `stereo_8_samples_48khz_s16le.wav`                   | Tiny test file. The first channel encodes 1, 2, ... 8. The second channel encodes 65535, 65534, ... 65528.          | 2        | 48kHz       | pcm_s16le | 8 samples
 `stereo_8_samples_48khz_s24le.wav`                   | Tiny test file. The first channel encodes 1, 2, ... 8. The second channel encodes 16777216, 16777215, ... 16777209. | 2        | 48kHz       | pcm_s24le | 8 samples
 `Transport_TOA_5s.wav`                               | Short clip of vehicles driving by using third-order ambisonics.                                                     | 16       | 48kHz       | pcm_s16le | 5s
+`Transport_9.1.6_5s.wav`                             | Short clip of vehicles driving by using 9.1.6.                                                                      | 16       | 48kHz       | pcm_s16le | 5s
+
+# Output WAV files
+
+Output wav files are based on the
+[layout](https://aomediacodec.github.io/iamf/#syntax-layout) in the mix
+presentation. Typically the ordering of channels is based on the related
+[ITU-2051-3](https://www.itu.int/rec/R-REC-BS.2051) layout.
+
+Mix Presentation Layout | Channel Order Convention | Channel Order
+----------------------- | ------------------------ | -------------
+Sound System A (0+2+0)  | ITU-2051-3               | L, R
+Sound System B (0+5+0)  | ITU-2051-3               | L, R, C, LFE, Ls, Rs
+Sound System C (2+5+0)  | ITU-2051-3               | L, R, C, LFE, Ls, Rs, Ltf, Rtf
+Sound System D (4+5+0)  | ITU-2051-3               | L, R, C, LFE, Ls, Rs, Ltf, Rtf, Ltr, Rtr
+Sound System E (4+5+1)  | ITU-2051-3               | L, R, C, LFE, Ls, Rs, Ltf, Rtf, Ltr, Rtr, Cbf
+Sound System F (3+7+0)  | ITU-2051-3               | C, L, R, LH, RH, LS, LB, RB, CH, LFE1, LFE2
+Sound System G (4+9+0)  | ITU-2051-3               | L, R, C, LFE, Lss, Rss, Lrs, Rrs, Ltf, Rtf, Ltb, Rtb, Lsc, Rsc
+Sound System H (9+10+3) | ITU-2051-3               | FL, FR, FC, LFE1, BL, BR, FLc, FRc, BC, LFE2, SiL, SiR, TpFL, TpFR, TpFC, TpC, TpBL, TpBR, TpSiL, TpSiR, TpBC, BtFC, BtFL, BtFR
+Sound System I (0+7+0)  | ITU_2051-3               | L, R, C, LFE, Lss, Rss, Lrs, Rrs
+Sound System J (4+7+0)  | ITU_2051-3               | L, R, C, LFE, Lss, Rss, Lrs, Rrs, Ltf, Rtf, Ltb, Rtb
+Sound System 10         | IAMF                     | L7, R7, C, LFE, Lss7, Rss7, Lrs7, Rrs7, Ltf2, Rtf2
+Sound System 11         | IAMF                     | L3, R3, C, LFE, Ltf3, Rtf3,
+Sound System 12         | IAMF                     | C
+Sound System 13         | IAMF                     | FL, FR, FC, LFE, BL, BR, FLc, FRc, SiL, SiR, TpFL, TpFR, TpBL, TpBR, TpSiL, TpSiR
+Binaural Layout         | IAMF                     | L2, R2