Merge branch 'main' of https://github.com/yilun-zhangs/libiamf into main

Author: yilun-zhangs

code/dep_external/include/resonance/iamf_resonance_api.h

@@ -153,7 +153,7 @@ EXPORT_API void* CreateResonanceAudioApi2(size_t num_channels,
 //  // class construction.
 // static const SourceId kInvalidSourceId = -1;
 
-EXPORT_API void DestoryResonanceAudioApi2(void* pv_thiz);
+EXPORT_API void DestroyResonanceAudioApi2(void* pv_thiz);
 
 // Renders and outputs an interleaved output buffer in float format.
 //

code/dep_external/lib/binaural/libiamf2resonance.so

@@ -38,7 +38,7 @@ namespace vraudio {
     // class construction.
     //static const SourceId ` = -1;
 
-    extern "C" EXPORT_API void DestoryResonanceAudioApi2(void* pv_thiz)
+    extern "C" EXPORT_API void DestroyResonanceAudioApi2(void* pv_thiz)
     {
         ResonanceAudioApi* thiz = (ResonanceAudioApi*)pv_thiz;
         delete thiz;

code/dep_external/src/binaural/iamf2resonance/iamf2resonance.cc

@@ -153,7 +153,7 @@ EXPORT_API void* CreateResonanceAudioApi2(size_t num_channels,
 //  // class construction.
 // static const SourceId kInvalidSourceId = -1;
 
-EXPORT_API void DestoryResonanceAudioApi2(void* pv_thiz);
+EXPORT_API void DestroyResonanceAudioApi2(void* pv_thiz);
 
 // Renders and outputs an interleaved output buffer in float format.
 //

code/dep_external/src/binaural/iamf2resonance/iamf_resonance_api.h

@@ -85,7 +85,7 @@ void IAMF_element_renderer_deinit_H2B(binaural_filter_t* binaural_f,
       }
     }
     if (i == N_SOURCE_ELM) {
-      DestoryResonanceAudioApi2(binaural_f->h2b_api);
+      DestroyResonanceAudioApi2(binaural_f->h2b_api);
       binaural_f->h2b_init = 0;
     }
   }

code/src/iamf_dec/h2b_rdr.c

@@ -50,7 +50,7 @@ int opus_multistream2_decode(OpusMS2Decoder *st, uint8_t *buffer[],
                              uint32_t len[], void *pcm, uint32_t frame_size);
 
 /**
- * @brief     Destory an opus decoder.
+ * @brief     Destroy an opus decoder.
  * @param     [in] st : opus decoder handle.
  */
 void opus_multistream2_decoder_destroy(OpusMS2Decoder *st);

code/src/iamf_dec/opus/opus_multistream2_decoder.h

@@ -54,11 +54,23 @@ enum InsertionHook {
   INSERTION_HOOK_INVALID = 0;
   INSERTION_HOOK_BEFORE_DESCRIPTORS = 1;
   INSERTION_HOOK_AFTER_DESCRIPTORS = 2;
+  INSERTION_HOOK_AFTER_IA_SEQUENCE_HEADER = 3;
+  INSERTION_HOOK_AFTER_CODEC_CONFIGS = 4;
+  INSERTION_HOOK_AFTER_AUDIO_ELEMENTS = 5;
+  INSERTION_HOOK_AFTER_MIX_PRESENTATIONS = 6;
+  INSERTION_HOOK_BEFORE_PARAMETER_BLOCKS_AT_TICK = 7;
+  INSERTION_HOOK_AFTER_PARAMETER_BLOCKS_AT_TICK = 8;
+  INSERTION_HOOK_AFTER_AUDIO_FRAMES_AT_TICK = 9;
 }
 
 message ArbitraryObuMetadata {
   optional InsertionHook insertion_hook = 1;
   optional ArbitraryObuType obu_type = 4;
   optional ObuHeaderMetadata obu_header = 2;
   optional bytes payload = 3;
+  // Used to force the bitstream to be invalid when this OBU is inserted.
+  optional bool invalidates_bitstream = 5;
+  // Used to control the temporal unit when using the *_AT_TICK insertion hooks.
+  // Ignored otherwise.
+  optional int64 insertion_tick = 6;
 }

proto/arbitrary_obu.proto

@@ -27,16 +27,36 @@ message ParamDefinitionExtension {
 }
 
 message AudioElementParam {
-  optional uint32 param_definition_type = 1;
+  optional uint32 deprecated_param_definition_type = 1 [deprecated = true];
+  optional ParamDefinitionType param_definition_type = 5;
   oneof param_definition {
     DemixingParamDefinition demixing_param = 2;
     ReconGainParamDefinition recon_gain_param = 3;
     ParamDefinitionExtension param_definition_extension = 4;
   }
 }
 
+// Valid proto enums start at index 1, which are different from the
+// corresponding enums in C++, e.g. kLayoutMono = 0.
+enum LoudspeakerLayout {
+  LOUDSPEAKER_LAYOUT_INVALID = 0;
+  LOUDSPEAKER_LAYOUT_MONO = 1;
+  LOUDSPEAKER_LAYOUT_STEREO = 2;
+  LOUDSPEAKER_LAYOUT_5_1_CH = 3;
+  LOUDSPEAKER_LAYOUT_5_1_2_CH = 4;
+  LOUDSPEAKER_LAYOUT_5_1_4_CH = 5;
+  LOUDSPEAKER_LAYOUT_7_1_CH = 6;
+  LOUDSPEAKER_LAYOUT_7_1_2_CH = 7;
+  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;
+}
+
 message ChannelAudioLayerConfig {
-  optional uint32 loudspeaker_layout = 1;
+  optional uint32 deprecated_loudspeaker_layout = 1 [deprecated = true];
+  optional LoudspeakerLayout loudspeaker_layout = 10;
   optional uint32 output_gain_is_present_flag = 2;
   optional uint32 recon_gain_is_present_flag = 3;
   optional uint32 reserved_a = 4;

proto/audio_element.proto

@@ -15,7 +15,7 @@ package libiamf_proto;
 import "obu_header.proto";
 
 // Valid proto enums start at index 1, which are different from the
-// corresponding enums in C, e.g. kLpcmBigEndian = 0.
+// corresponding enums in C++, e.g. kLpcmBigEndian = 0.
 enum LpcmFormatFlags {
   LPCM_INVALID = 0;
   LPCM_BIG_ENDIAN = 1;
@@ -39,19 +39,21 @@ enum OpusApplicationFlag {
 message OpusEncoderMetadata {
   optional int32 target_bitrate_per_channel = 1;
   optional OpusApplicationFlag application = 2;
+  optional bool use_float_api = 3 [default = true];
+  optional float coupling_rate_adjustment = 4 [default = 1.0];
 }
 
 message OpusDecoderConfig {
   optional uint32 version = 1;
-  optional uint32 output_channel_count = 2;
+  optional uint32 output_channel_count = 2 [default = 2];
   optional uint32 pre_skip = 3;
   optional uint32 input_sample_rate = 4;
-  optional int32 output_gain = 5;
-  optional uint32 mapping_family = 6;
+  optional int32 output_gain = 5 [default = 0];
+  optional uint32 mapping_family = 6 [default = 0];
   optional OpusEncoderMetadata opus_encoder_metadata = 7;
 }
 
-enum AacSampleFrequencyIndex {
+enum SampleFrequencyIndex {
   AAC_SAMPLE_FREQUENCY_INDEX_INVALID = 0;
   AAC_SAMPLE_FREQUENCY_INDEX_96000 = 1;
   AAC_SAMPLE_FREQUENCY_INDEX_88200 = 2;
@@ -80,23 +82,23 @@ message AacEncoderMetadata {
 
 message AacDecoderSpecificInfo {
   optional uint32 decoder_specific_info_descriptor_tag = 5 [default = 0x05];
-  optional uint32 audio_object_type = 1;
-  optional AacSampleFrequencyIndex sample_frequency_index = 2;
+  optional uint32 audio_object_type = 1 [default = 2];
+  optional SampleFrequencyIndex sample_frequency_index = 2;
   optional uint32 sampling_frequency = 3;
-  optional uint32 channel_configuration = 4;
+  optional uint32 channel_configuration = 4 [default = 2];
 }
 
 message AacGaSpecificConfig {
-  optional bool frame_length_flag = 1;
-  optional bool depends_on_core_coder = 2;
-  optional bool extension_flag = 3;
+  optional bool frame_length_flag = 1 [default = false];
+  optional bool depends_on_core_coder = 2 [default = false];
+  optional bool extension_flag = 3 [default = false];
 }
 
 message AacDecoderConfig {
   optional uint32 decoder_config_descriptor_tag = 11 [default = 0x04];
-  optional uint32 object_type_indication = 1;
-  optional uint32 stream_type = 2;
-  optional bool upstream = 3;
+  optional uint32 object_type_indication = 1 [default = 0x40];
+  optional uint32 stream_type = 2 [default = 5];
+  optional bool upstream = 3 [default = false];
   optional bool reserved = 4;
   optional uint32 buffer_size_db = 5;
   optional uint32 max_bitrate = 6;
@@ -126,13 +128,15 @@ message FlacMetaBlockHeader {
 message FlacMetaBlockStreamInfo {
   optional uint32 minimum_block_size = 1;
   optional uint32 maximum_block_size = 2;
-  optional uint32 minimum_frame_size = 3;
-  optional uint32 maximum_frame_size = 4;
+  optional uint32 minimum_frame_size = 3 [default = 0];
+  optional uint32 maximum_frame_size = 4 [default = 0];
   optional uint32 sample_rate = 5;
-  optional uint32 number_of_channels = 6;
+  optional uint32 number_of_channels = 6 [default = 1];
   optional uint32 bits_per_sample = 7;
   optional uint64 total_samples_in_stream = 8;
-  optional bytes md5_signature = 9;
+  optional bytes md5_signature = 9
+      [default =
+           "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00"];
 }
 
 // Settings to configure `libflac`.
@@ -153,10 +157,19 @@ message FlacDecoderConfig {
   optional FlacEncoderMetadata flac_encoder_metadata = 4;
 }
 
+enum CodecId {
+  CODEC_ID_INVALID = 0;
+  CODEC_ID_OPUS = 0x4f707573;    // "Opus"
+  CODEC_ID_LPCM = 0x6970636d;    // "ipcm"
+  CODEC_ID_AAC_LC = 0x6d703461;  // "mp4a"
+  CODEC_ID_FLAC = 0x664c6143;    // "fLaC"
+}
+
 message CodecConfig {
-  optional uint32 codec_id = 1;
+  optional uint32 deprecated_codec_id = 1 [deprecated = true];
+  optional CodecId codec_id = 10;
   optional uint32 num_samples_per_frame = 2;
-  reserved 3;
+  reserved 3, 4;
   optional int32 audio_roll_distance = 9;
   oneof decoder_config {
     LpcmDecoderConfig decoder_config_lpcm = 5;

proto/codec_config.proto

@@ -14,15 +14,16 @@ package libiamf_proto;
 
 import "obu_header.proto";
 
-enum ProfileVerson {
+enum ProfileVersion {
   PROFILE_VERSION_INVALID = 0;
   PROFILE_VERSION_SIMPLE = 1;
   PROFILE_VERSION_BASE = 2;
+  PROFILE_VERSION_BASE_ENHANCED = 3;
 }
 
 message IASequenceHeaderObuMetadata {
-  optional uint32 ia_code = 1;
-  optional ProfileVerson primary_profile = 2;
-  optional ProfileVerson additional_profile = 3;
+  optional uint32 ia_code = 1 [default = 0x69616d66 /* "iamf" */];
+  optional ProfileVersion primary_profile = 2;
+  optional ProfileVersion additional_profile = 3;
   optional ObuHeaderMetadata obu_header = 4;
 }