public void SendAudioPacket(string requestId, byte[] data)
{
byte[] headerBytes;
byte[] headerHead;
headerBytes = BuildAudioPacketHeader(requestId);
headerHead = BuildAudioPacketHeaderHead(headerBytes);
// The WebSocket Speech protocol docs state that PCM audio must be sampled at 16 kHz with 16 bits per sample
// and one channel (riff-16khz-16bit-mono-pcm) but 48kHz-16-bit-stereo works too, more testing is required.
var arr = headerHead.Concat(headerBytes).Concat(data).ToArray();
var arrSeg = new ArraySegment <byte>(arr, 0, arr.Length);
Debug.Log($"Sending audio data sample from microphone.");
if (SpeechWebSocketClient.State != WebSocketState.Open)
{
return;
}
SpeechWebSocketClient.SendAsync(arrSeg, WebSocketMessageType.Binary, true, new CancellationToken());
Debug.Log($"Audio data packet from microphone sent successfully!");
//var dt = Encoding.ASCII.GetString(arr);
}
public async Task <bool> CreateSpeechRecognitionJobFromFile(string audioFilePath, string authenticationToken, string region)
{
try
{
state = JobState.PreparingJob;
SpeechWebSocketClient = await InitializeSpeechWebSocketClient(authenticationToken, region);
var receiving = Receiving(SpeechWebSocketClient);
var sending = Task.Run(async() =>
{
// Create a unique request ID, must be a UUID in "no-dash" format
var requestId = Guid.NewGuid().ToString("N");
ArraySegment <byte> buffer = CreateSpeechConfigMessagePayloadBuffer(requestId);
if (SpeechWebSocketClient.State != WebSocketState.Open)
{
return;
}
Debug.Log("Sending speech.config...");
// Send speech.config to Speech Service
await SpeechWebSocketClient.SendAsync(buffer, WebSocketMessageType.Text, true, new CancellationToken());
Debug.Log("speech.config sent successfully!");
// SENDING AUDIO TO SPEECH SERVICE:
// Speech-enabled client applications send audio to Speech Service by converting the audio stream
// into a series of audio chunks. Each chunk of audio carries a segment of the spoken audio that's
// to be transcribed by the service. The maximum size of a single audio chunk is 8,192 bytes.
// Audio stream messages are Binary WebSocket messages.
Debug.Log($"Preparing to send audio file: {audioFilePath}");
FileInfo audioFileInfo = new FileInfo(audioFilePath);
FileStream audioFileStream = audioFileInfo.OpenRead();
state = JobState.ProcessingAudio;
byte[] headerBytes;
byte[] headerHead;
for (int cursor = 0; cursor < audioFileInfo.Length; cursor++)
{
headerBytes = BuildAudioPacketHeader(requestId);
headerHead = BuildAudioPacketHeaderHead(headerBytes);
// PCM audio must be sampled at 16 kHz with 16 bits per sample and one channel (riff-16khz-16bit-mono-pcm).
var byteLen = 8192 - headerBytes.Length - 2;
var fbuff = new byte[byteLen];
audioFileStream.Read(fbuff, 0, byteLen);
var arr = headerHead.Concat(headerBytes).Concat(fbuff).ToArray();
var arrSeg = new ArraySegment <byte>(arr, 0, arr.Length);
Debug.Log($"Sending audio data from position: {cursor}");
if (SpeechWebSocketClient.State != WebSocketState.Open)
{
return;
}
cursor += byteLen;
var end = cursor >= audioFileInfo.Length;
await SpeechWebSocketClient.SendAsync(arrSeg, WebSocketMessageType.Binary, true, new CancellationToken());
Debug.Log($"Audio data from file {audioFilePath} sent successfully!");
var dt = Encoding.ASCII.GetString(arr);
}
await SendEmptyAudioMessageToWebSocketClient(SpeechWebSocketClient, requestId);
audioFileStream.Dispose();
});
// Wait for tasks to complete
await Task.WhenAll(sending, receiving);
if (sending.IsFaulted)
{
state = JobState.Error;
var err = sending.Exception;
throw err;
}
if (receiving.IsFaulted)
{
state = JobState.Error;
var err = receiving.Exception;
throw err;
}
return(true);
}
catch (Exception ex)
{
state = JobState.Error;
Debug.Log($"An exception occurred during creation of Speech Recognition job from audio file {audioFilePath}:"
+ Environment.NewLine + ex.Message);
return(false);
}
}
/// <summary>
/// prepares a new speech recognition job, sending the proper headers including the audio data header
/// </summary>
/// <param name="authenticationToken"></param>
/// <param name="region"></param>
/// <param name="resolution"></param>
/// <param name="channels"></param>
/// <param name="rate"></param>
/// <returns></returns>
public async Task <bool> CreateSpeechRecognitionJobFromVoice(string authenticationToken, string region, int resolution, int channels, int rate)
{
try
{
state = JobState.PreparingJob;
SpeechWebSocketClient = await InitializeSpeechWebSocketClient(authenticationToken, region);
var receiving = Receiving(SpeechWebSocketClient);
var sending = Task.Run(async() =>
{
// Create a unique request ID, must be a UUID in "no-dash" format
CurrentRequestId = Guid.NewGuid().ToString("N");
ArraySegment <byte> buffer = CreateSpeechConfigMessagePayloadBuffer(CurrentRequestId);
if (SpeechWebSocketClient.State != WebSocketState.Open)
{
return;
}
Debug.Log("Sending speech.config...");
// Send speech.config to Speech Service
await SpeechWebSocketClient.SendAsync(buffer, WebSocketMessageType.Text, true, new CancellationToken());
Debug.Log("speech.config sent successfully!");
// SENDING AUDIO TO SPEECH SERVICE:
// Speech-enabled client applications send audio to Speech Service by converting the audio stream
// into a series of audio chunks. Each chunk of audio carries a segment of the spoken audio that's
// to be transcribed by the service. The maximum size of a single audio chunk is 8,192 bytes.
// Audio stream messages are Binary WebSocket messages.
// First we need to send an audio packet with the RIFF PCM (WAV) data header. Note that we don't know how
// many samples we'll have since we are recording live, so we set nbsamples to zero.
// The WebSocket Speech protocol docs state that PCM audio must be sampled at 16 kHz with 16 bits per sample
// and one channel (riff-16khz-16bit-mono-pcm) but 48kHz-16-bit-stereo works too, more testing is required.
var wavHeader = BuildRiffWAVHeader(0, resolution, channels, rate);
SendAudioPacket(CurrentRequestId, wavHeader);
Debug.Log($"First Audio data paket with WAV header sent successfully!");
Debug.Log($"WebSocket Client is now ready to receive audio packets from the microphone.");
state = JobState.ReadyForAudioPackets;
});
// Wait for tasks to complete
await Task.WhenAll(sending, receiving);
if (sending.IsFaulted)
{
var err = sending.Exception;
throw err;
}
if (receiving.IsFaulted)
{
var err = receiving.Exception;
throw err;
}
return(true);
}
catch (Exception ex)
{
Debug.Log($"An exception occurred during creation of Speech Recognition job from microphone:"
+ Environment.NewLine + ex.Message);
return(false);
}
}