record audio duration and encoding in tacotron2 job

crates/lib/mysql_queries/src/queries/media_files/create/insert_media_file_from_tacotron2.rs

@@ -20,11 +20,9 @@ pub struct InsertMediaFileArgs<'a> {
 
   pub maybe_mime_type: Option<&'a str>,
   pub file_size_bytes: u64,
-  pub duration_millis: u64,
   pub sha256_checksum: &'a str,
-
-  // TODO: Media duration.
-  //pub duration_millis: u64,
+  pub maybe_duration_millis: Option<u64>,
+  pub maybe_audio_codec_name: Option<&'a str>,
 
   pub public_bucket_directory_hash: &'a str,
   pub maybe_public_bucket_prefix: Option<&'a str>,
@@ -90,6 +88,9 @@ SET
   maybe_mime_type = ?,
   file_size_bytes = ?,
 
+  maybe_duration_millis = ?,
+  maybe_audio_encoding = ?,
+
   checksum_sha2 = ?,
 
   public_bucket_directory_hash = ?,
@@ -124,6 +125,9 @@ SET
       args.maybe_mime_type,
       args.file_size_bytes,
 
+      args.maybe_duration_millis,
+      args.maybe_audio_codec_name,
+
       args.sha256_checksum,
 
       args.public_bucket_directory_hash,

crates/service/job/inference_job/src/job/job_types/tts/tacotron2_v2_early_fakeyou/process_job.rs

@@ -11,6 +11,7 @@ use errors::AnyhowResult;
 use filesys::check_file_exists::check_file_exists;
 use filesys::safe_delete_temp_directory::safe_delete_temp_directory;
 use filesys::safe_delete_temp_file::safe_delete_temp_file;
+use media::decode_basic_audio_info::decode_basic_audio_file_info;
 use migration::text_to_speech::get_tts_model_for_run_inference_migration::TtsModelForRunInferenceMigrationWrapper;
 use mysql_queries::column_types::vocoder_type::VocoderType;
 use mysql_queries::queries::generic_inference::job::list_available_generic_inference_jobs::AvailableInferenceJob;
@@ -309,6 +310,17 @@ async fn process_job_with_cleanup(
 
   safe_delete_temp_file(&output_metadata_fs_path);
 
+  info!("Decoding audio info...");
+
+  // NB: The metadata file emitted by tacotron2 hasn't been recording the audio duration recently. (Perhaps it never worked?)
+  let mut maybe_audio_duration_millis = file_metadata.duration_millis.clone();
+  let mut maybe_audio_codec_name = None;
+
+  if let Ok(audio_info) = decode_basic_audio_file_info(&output_audio_fs_path, Some("audio/wav"), None) {
+    maybe_audio_duration_millis = audio_info.duration_millis;
+    maybe_audio_codec_name = audio_info.codec_name;
+  }
+
   // ==================== UPLOAD FILES AND SAVE RECORD ==================== //
 
   let inference_result = upload_results(UploadResultArgs {
@@ -319,6 +331,8 @@ async fn process_job_with_cleanup(
     work_temp_dir,
     pretrained_vocoder,
     file_metadata: &file_metadata,
+    maybe_audio_duration_millis,
+    maybe_audio_codec_name: maybe_audio_codec_name.as_deref(),
     output_audio_fs_path: &output_audio_fs_path,
     output_spectrogram_fs_path: &output_spectrogram_fs_path,
     upload_as_media_file: model_dependencies.upload_as_media_file,

crates/service/job/inference_job/src/job/job_types/tts/tacotron2_v2_early_fakeyou/upload_results.rs

@@ -38,6 +38,8 @@ pub struct UploadResultArgs<'a> {
 
   // Outputs
   pub file_metadata: &'a FileMetadata,
+  pub maybe_audio_duration_millis: Option<u64>,
+  pub maybe_audio_codec_name: Option<&'a str>,
   pub output_audio_fs_path: &'a Path,
   pub output_spectrogram_fs_path: &'a Path,
 
@@ -107,7 +109,8 @@ async fn upload_as_media_file(args: UploadResultArgs<'_>) -> Result<ResultDetail
     job: &args.job,
     maybe_mime_type: Some(&MIME_TYPE),
     file_size_bytes: args.file_metadata.file_size_bytes,
-    duration_millis: args.file_metadata.duration_millis.unwrap_or(0),
+    maybe_duration_millis: args.maybe_audio_duration_millis,
+    maybe_audio_codec_name: args.maybe_audio_codec_name,
     sha256_checksum: &file_checksum,
     public_bucket_directory_hash: result_bucket_location.get_object_hash(),
     maybe_public_bucket_prefix: Some(BUCKET_FILE_PREFIX),