public LongRunningRecognizeResponse TranscribeInCloud(TranscribeParameters transParams)
{
// var speechClient = SpeechClient.Create();
string fileOnCloudStorage = "gs://" + transParams.GoogleCloudBucketName + "/" + transParams.objectName;
RecognitionAudio recogAudio = RecognitionAudio.FromStorageUri(fileOnCloudStorage);
SpeakerDiarizationConfig sdc = new SpeakerDiarizationConfig()
{
EnableSpeakerDiarization = true,
MinSpeakerCount = transParams.MinSpeakerCount,
MaxSpeakerCount = transParams.MaxSpeakerCount
};
var longOperation = speechClient.LongRunningRecognize(new RecognitionConfig()
{
Encoding = RecognitionConfig.Types.AudioEncoding.Flac,
SampleRateHertz = 44100,
EnableWordTimeOffsets = true,
LanguageCode = transParams.language,
EnableAutomaticPunctuation = true,
DiarizationConfig = sdc,
SpeechContexts =
{
new SpeechContext {
Phrases = { transParams.phrases}
}
}
}, recogAudio);
longOperation = longOperation.PollUntilCompleted();
var response = longOperation.Result;
return(response);
}
public LongRunningRecognizeResponse UploadAndTranscribeInCloud(TranscribeParameters transParams)
{
UploadToCloudIfNeeded(transParams);
LongRunningRecognizeResponse response = TranscribeInCloud(transParams);
return(response);
}
private void UploadToCloudIfNeeded(TranscribeParameters transParams)
{
GoogleBucket gb = new GoogleBucket();
if (transParams.useAudioFileAlreadyInCloud)
{
// Only upload if not in cloud
if (!gb.IsObjectInBucket(transParams.GoogleCloudBucketName, transParams.objectName))
{
gb.UploadFile(transParams.GoogleCloudBucketName, transParams.audiofilePath, transParams.objectName, "audio /x-flac");
}
}
}
public Transcribed_Dto TranscribeAudioFile(TranscribeParameters transParams, string rawResponseFile = null)
{
LongRunningRecognizeResponse response = UploadAndTranscribeInCloud(transParams);
// Save the raw response, if we were passed a file path.
if (rawResponseFile != "")
{
string responseString = JsonConvert.SerializeObject(response, Formatting.Indented);
File.WriteAllText(rawResponseFile, responseString);
}
Transcribed_Dto resp = TransformResponse.Simpify(response.Results);
return(TransformResponse.FixSpeakerTags(resp));
}
