Transcribing audio from streaming input

This section demonstrates how to transcribe streaming audio, like the input from a microphone, to text.

Streaming speech recognition allows you to stream audio to Speech-to-Text and receive a stream speech recognition results in real time as the audio is processed. See also the audio limits for streaming speech recognition requests. Streaming speech recognition is available via gRPC only.

Performing streaming speech recognition on a local file

Below is an example of performing streaming speech recognition on a local audio file. There is a 10 MB limit on all streaming requests sent to the API. This limit applies to to both the initial StreamingRecognize request and the size of each individual message in the stream. Exceeding this limit will throw an error.

C#

static async Task<object> StreamingRecognizeAsync(string filePath)
{
    var speech = SpeechClient.Create();
    var streamingCall = speech.StreamingRecognize();
    // Write the initial request with the config.
    await streamingCall.WriteAsync(
        new StreamingRecognizeRequest()
        {
            StreamingConfig = new StreamingRecognitionConfig()
            {
                Config = new RecognitionConfig()
                {
                    Encoding =
                    RecognitionConfig.Types.AudioEncoding.Linear16,
                    SampleRateHertz = 16000,
                    LanguageCode = "en",
                },
                InterimResults = true,
            }
        });
    // Print responses as they arrive.
    Task printResponses = Task.Run(async () =>
    {
        var responseStream = streamingCall.GetResponseStream();
        while (await responseStream.MoveNextAsync())
        {
            StreamingRecognizeResponse response = responseStream.Current;
            foreach (StreamingRecognitionResult result in response.Results)
            {
                foreach (SpeechRecognitionAlternative alternative in result.Alternatives)
                {
                    Console.WriteLine(alternative.Transcript);
                }
            }
        }
    });
    // Stream the file content to the API.  Write 2 32kb chunks per
    // second.
    using (FileStream fileStream = new FileStream(filePath, FileMode.Open))
    {
        var buffer = new byte[32 * 1024];
        int bytesRead;
        while ((bytesRead = await fileStream.ReadAsync(
            buffer, 0, buffer.Length)) > 0)
        {
            await streamingCall.WriteAsync(
                new StreamingRecognizeRequest()
                {
                    AudioContent = Google.Protobuf.ByteString
                    .CopyFrom(buffer, 0, bytesRead),
                });
            await Task.Delay(500);
        };
    }
    await streamingCall.WriteCompleteAsync();
    await printResponses;
    return 0;
}

Go

import (
	"context"
	"flag"
	"fmt"
	"io"
	"log"
	"os"
	"path/filepath"

	speech "cloud.google.com/go/speech/apiv1"
	speechpb "google.golang.org/genproto/googleapis/cloud/speech/v1"
)

func main() {
	flag.Usage = func() {
		fmt.Fprintf(os.Stderr, "Usage: %s <AUDIOFILE>\n", filepath.Base(os.Args[0]))
		fmt.Fprintf(os.Stderr, "<AUDIOFILE> must be a path to a local audio file. Audio file must be a 16-bit signed little-endian encoded with a sample rate of 16000.\n")

	}
	flag.Parse()
	if len(flag.Args()) != 1 {
		log.Fatal("Please pass path to your local audio file as a command line argument")
	}
	audioFile := flag.Arg(0)

	ctx := context.Background()

	client, err := speech.NewClient(ctx)
	if err != nil {
		log.Fatal(err)
	}
	stream, err := client.StreamingRecognize(ctx)
	if err != nil {
		log.Fatal(err)
	}
	// Send the initial configuration message.
	if err := stream.Send(&speechpb.StreamingRecognizeRequest{
		StreamingRequest: &speechpb.StreamingRecognizeRequest_StreamingConfig{
			StreamingConfig: &speechpb.StreamingRecognitionConfig{
				Config: &speechpb.RecognitionConfig{
					Encoding:        speechpb.RecognitionConfig_LINEAR16,
					SampleRateHertz: 16000,
					LanguageCode:    "en-US",
				},
			},
		},
	}); err != nil {
		log.Fatal(err)
	}

	f, err := os.Open(audioFile)
	if err != nil {
		log.Fatal(err)
	}
	defer f.Close()

	go func() {
		buf := make([]byte, 1024)
		for {
			n, err := f.Read(buf)
			if n > 0 {
				if err := stream.Send(&speechpb.StreamingRecognizeRequest{
					StreamingRequest: &speechpb.StreamingRecognizeRequest_AudioContent{
						AudioContent: buf[:n],
					},
				}); err != nil {
					log.Printf("Could not send audio: %v", err)
				}
			}
			if err == io.EOF {
				// Nothing else to pipe, close the stream.
				if err := stream.CloseSend(); err != nil {
					log.Fatalf("Could not close stream: %v", err)
				}
				return
			}
			if err != nil {
				log.Printf("Could not read from %s: %v", audioFile, err)
				continue
			}
		}
	}()

	for {
		resp, err := stream.Recv()
		if err == io.EOF {
			break
		}
		if err != nil {
			log.Fatalf("Cannot stream results: %v", err)
		}
		if err := resp.Error; err != nil {
			log.Fatalf("Could not recognize: %v", err)
		}
		for _, result := range resp.Results {
			fmt.Printf("Result: %+v\n", result)
		}
	}
}

Java

/**
 * Performs streaming speech recognition on raw PCM audio data.
 *
 * @param fileName the path to a PCM audio file to transcribe.
 */
public static void streamingRecognizeFile(String fileName) throws Exception, IOException {
  Path path = Paths.get(fileName);
  byte[] data = Files.readAllBytes(path);

  // Instantiates a client with GOOGLE_APPLICATION_CREDENTIALS
  try (SpeechClient speech = SpeechClient.create()) {

    // Configure request with local raw PCM audio
    RecognitionConfig recConfig =
        RecognitionConfig.newBuilder()
            .setEncoding(AudioEncoding.LINEAR16)
            .setLanguageCode("en-US")
            .setSampleRateHertz(16000)
            .setModel("default")
            .build();
    StreamingRecognitionConfig config =
        StreamingRecognitionConfig.newBuilder().setConfig(recConfig).build();

    class ResponseApiStreamingObserver<T> implements ApiStreamObserver<T> {
      private final SettableFuture<List<T>> future = SettableFuture.create();
      private final List<T> messages = new java.util.ArrayList<T>();

      @Override
      public void onNext(T message) {
        messages.add(message);
      }

      @Override
      public void onError(Throwable t) {
        future.setException(t);
      }

      @Override
      public void onCompleted() {
        future.set(messages);
      }

      // Returns the SettableFuture object to get received messages / exceptions.
      public SettableFuture<List<T>> future() {
        return future;
      }
    }

    ResponseApiStreamingObserver<StreamingRecognizeResponse> responseObserver =
        new ResponseApiStreamingObserver<>();

    BidiStreamingCallable<StreamingRecognizeRequest, StreamingRecognizeResponse> callable =
        speech.streamingRecognizeCallable();

    ApiStreamObserver<StreamingRecognizeRequest> requestObserver =
        callable.bidiStreamingCall(responseObserver);

    // The first request must **only** contain the audio configuration:
    requestObserver.onNext(
        StreamingRecognizeRequest.newBuilder().setStreamingConfig(config).build());

    // Subsequent requests must **only** contain the audio data.
    requestObserver.onNext(
        StreamingRecognizeRequest.newBuilder()
            .setAudioContent(ByteString.copyFrom(data))
            .build());

    // Mark transmission as completed after sending the data.
    requestObserver.onCompleted();

    List<StreamingRecognizeResponse> responses = responseObserver.future().get();

    for (StreamingRecognizeResponse response : responses) {
      // For streaming recognize, the results list has one is_final result (if available) followed
      // by a number of in-progress results (if iterim_results is true) for subsequent utterances.
      // Just print the first result here.
      StreamingRecognitionResult result = response.getResultsList().get(0);
      // There can be several alternative transcripts for a given chunk of speech. Just use the
      // first (most likely) one here.
      SpeechRecognitionAlternative alternative = result.getAlternativesList().get(0);
      System.out.printf("Transcript : %s\n", alternative.getTranscript());
    }
  }
}

Node.js

const fs = require('fs');

// Imports the Google Cloud client library
const speech = require('@google-cloud/speech');

// Creates a client
const client = new speech.SpeechClient();

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const filename = 'Local path to audio file, e.g. /path/to/audio.raw';
// const encoding = 'Encoding of the audio file, e.g. LINEAR16';
// const sampleRateHertz = 16000;
// const languageCode = 'BCP-47 language code, e.g. en-US';

const request = {
  config: {
    encoding: encoding,
    sampleRateHertz: sampleRateHertz,
    languageCode: languageCode,
  },
  interimResults: false, // If you want interim results, set this to true
};

// Stream the audio to the Google Cloud Speech API
const recognizeStream = client
  .streamingRecognize(request)
  .on('error', console.error)
  .on('data', data => {
    console.log(
      `Transcription: ${data.results[0].alternatives[0].transcript}`
    );
  });

// Stream an audio file from disk to the Speech API, e.g. "./resources/audio.raw"
fs.createReadStream(filename).pipe(recognizeStream);

PHP

use Google\Cloud\Speech\V1\SpeechClient;
use Google\Cloud\Speech\V1\RecognitionConfig;
use Google\Cloud\Speech\V1\StreamingRecognitionConfig;
use Google\Cloud\Speech\V1\StreamingRecognizeRequest;
use Google\Cloud\Speech\V1\RecognitionConfig\AudioEncoding;

/** Uncomment and populate these variables in your code */
// $audioFile = 'path to an audio file';

// change these variables if necessary
$encoding = AudioEncoding::LINEAR16;
$sampleRateHertz = 32000;
$languageCode = 'en-US';

// the gRPC extension is required for streaming
if (!extension_loaded('grpc')) {
    throw new \Exception('Install the grpc extension (pecl install grpc)');
}

$speechClient = new SpeechClient();
try {
    $config = (new RecognitionConfig())
        ->setEncoding($encoding)
        ->setSampleRateHertz($sampleRateHertz)
        ->setLanguageCode($languageCode);

    $strmConfig = new StreamingRecognitionConfig();
    $strmConfig->setConfig($config);

    $strmReq = new StreamingRecognizeRequest();
    $strmReq->setStreamingConfig($strmConfig);

    $strm = $speechClient->streamingRecognize();
    $strm->write($strmReq);

    $strmReq = new StreamingRecognizeRequest();
    $content = file_get_contents($audioFile);
    $strmReq->setAudioContent($content);
    $strm->write($strmReq);

    foreach ($strm->closeWriteAndReadAll() as $response) {
        foreach ($response->getResults() as $result) {
            foreach ($result->getAlternatives() as $alt) {
                printf("Transcription: %s\n", $alt->getTranscript());
            }
        }
    }
} finally {
    $speechClient->close();
}

Python

def transcribe_streaming(stream_file):
    """Streams transcription of the given audio file."""
    import io
    from google.cloud import speech

    client = speech.SpeechClient()

    with io.open(stream_file, "rb") as audio_file:
        content = audio_file.read()

    # In practice, stream should be a generator yielding chunks of audio data.
    stream = [content]

    requests = (speech.StreamingRecognizeRequest(audio_content=chunk)
                for chunk in stream)

    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=16000,

        language_code='en-US')

    streaming_config = speech.StreamingRecognitionConfig(config=config)

    # streaming_recognize returns a generator.
    responses = client.streaming_recognize(config=streaming_config, requests=requests,)

    for response in responses:
        # Once the transcription has settled, the first result will contain the
        # is_final result. The other results will be for subsequent portions of
        # the audio.
        for result in response.results:
            print("Finished: {}".format(result.is_final))
            print("Stability: {}".format(result.stability))
            alternatives = result.alternatives
            # The alternatives are ordered from most likely to least.
            for alternative in alternatives:
                print("Confidence: {}".format(alternative.confidence))
                print(u"Transcript: {}".format(alternative.transcript))

Ruby

# audio_file_path = "Path to file on which to perform speech recognition"

require "google/cloud/speech"

speech = Google::Cloud::Speech.speech

audio_content = File.binread audio_file_path
bytes_total   = audio_content.size
bytes_sent    = 0
chunk_size    = 32_000

input_stream = Gapic::StreamInput.new
output_stream = speech.streaming_recognize input_stream

config = {
  config: {
    encoding:                 :LINEAR16,
    sample_rate_hertz:        16_000,
    language_code:            "en-US",
    enable_word_time_offsets: true
  }
}
input_stream.push streaming_config: config

# Simulated streaming from a microphone
# Stream bytes...
while bytes_sent < bytes_total
  input_stream.push audio_content: audio_content[bytes_sent, chunk_size]
  bytes_sent += chunk_size
  sleep 1
end

puts "Stopped passing"
input_stream.close

results = output_stream

results.each do |result|
  puts "Transcript: #{result}"
end

While you can stream a local audio file to the Speech-to-Text API, it is recommended that you perform synchronous or asynchronous audio recognition for batch mode results.

Performing streaming speech recognition on an audio stream

Speech-to-Text can also perform recognition on streaming, real-time audio.

Here is an example of performing streaming speech recognition on an audio stream received from a microphone:

C#

static async Task<object> StreamingMicRecognizeAsync(int seconds)
{
    var speech = SpeechClient.Create();
    var streamingCall = speech.StreamingRecognize();
    // Write the initial request with the config.
    await streamingCall.WriteAsync(
        new StreamingRecognizeRequest()
        {
            StreamingConfig = new StreamingRecognitionConfig()
            {
                Config = new RecognitionConfig()
                {
                    Encoding =
                    RecognitionConfig.Types.AudioEncoding.Linear16,
                    SampleRateHertz = 16000,
                    LanguageCode = "en",
                },
                InterimResults = true,
            }
        });
    // Print responses as they arrive.
    Task printResponses = Task.Run(async () =>
    {
        var responseStream = streamingCall.GetResponseStream();
        while (await responseStream.MoveNextAsync())
        {
            StreamingRecognizeResponse response = responseStream.Current;
            foreach (StreamingRecognitionResult result in response.Results)
            {
                foreach (SpeechRecognitionAlternative alternative in result.Alternatives)
                {
                    Console.WriteLine(alternative.Transcript);
                }
            }
        }
    });
    // Read from the microphone and stream to API.
    object writeLock = new object();
    bool writeMore = true;
    var waveIn = new NAudio.Wave.WaveInEvent();
    waveIn.DeviceNumber = 0;
    waveIn.WaveFormat = new NAudio.Wave.WaveFormat(16000, 1);
    waveIn.DataAvailable +=
        (object sender, NAudio.Wave.WaveInEventArgs args) =>
        {
            lock (writeLock)
            {
                if (!writeMore)
                {
                    return;
                }

                streamingCall.WriteAsync(
                    new StreamingRecognizeRequest()
                    {
                        AudioContent = Google.Protobuf.ByteString
                            .CopyFrom(args.Buffer, 0, args.BytesRecorded)
                    }).Wait();
            }
        };
    waveIn.StartRecording();
    Console.WriteLine("Speak now.");
    await Task.Delay(TimeSpan.FromSeconds(seconds));
    // Stop recording and shut down.
    waveIn.StopRecording();
    lock (writeLock)
    {
        writeMore = false;
    }

    await streamingCall.WriteCompleteAsync();
    await printResponses;
    return 0;
}

Go

import (
	"context"
	"fmt"
	"io"
	"log"
	"os"

	speech "cloud.google.com/go/speech/apiv1"
	speechpb "google.golang.org/genproto/googleapis/cloud/speech/v1"
)

func main() {
	ctx := context.Background()

	client, err := speech.NewClient(ctx)
	if err != nil {
		log.Fatal(err)
	}
	stream, err := client.StreamingRecognize(ctx)
	if err != nil {
		log.Fatal(err)
	}
	// Send the initial configuration message.
	if err := stream.Send(&speechpb.StreamingRecognizeRequest{
		StreamingRequest: &speechpb.StreamingRecognizeRequest_StreamingConfig{
			StreamingConfig: &speechpb.StreamingRecognitionConfig{
				Config: &speechpb.RecognitionConfig{
					Encoding:        speechpb.RecognitionConfig_LINEAR16,
					SampleRateHertz: 16000,
					LanguageCode:    "en-US",
				},
			},
		},
	}); err != nil {
		log.Fatal(err)
	}

	go func() {
		// Pipe stdin to the API.
		buf := make([]byte, 1024)
		for {
			n, err := os.Stdin.Read(buf)
			if n > 0 {
				if err := stream.Send(&speechpb.StreamingRecognizeRequest{
					StreamingRequest: &speechpb.StreamingRecognizeRequest_AudioContent{
						AudioContent: buf[:n],
					},
				}); err != nil {
					log.Printf("Could not send audio: %v", err)
				}
			}
			if err == io.EOF {
				// Nothing else to pipe, close the stream.
				if err := stream.CloseSend(); err != nil {
					log.Fatalf("Could not close stream: %v", err)
				}
				return
			}
			if err != nil {
				log.Printf("Could not read from stdin: %v", err)
				continue
			}
		}
	}()

	for {
		resp, err := stream.Recv()
		if err == io.EOF {
			break
		}
		if err != nil {
			log.Fatalf("Cannot stream results: %v", err)
		}
		if err := resp.Error; err != nil {
			// Workaround while the API doesn't give a more informative error.
			if err.Code == 3 || err.Code == 11 {
				log.Print("WARNING: Speech recognition request exceeded limit of 60 seconds.")
			}
			log.Fatalf("Could not recognize: %v", err)
		}
		for _, result := range resp.Results {
			fmt.Printf("Result: %+v\n", result)
		}
	}
}

Java

/** Performs microphone streaming speech recognition with a duration of 1 minute. */
public static void streamingMicRecognize() throws Exception {

  ResponseObserver<StreamingRecognizeResponse> responseObserver = null;
  try (SpeechClient client = SpeechClient.create()) {

    responseObserver =
        new ResponseObserver<StreamingRecognizeResponse>() {
          ArrayList<StreamingRecognizeResponse> responses = new ArrayList<>();

          public void onStart(StreamController controller) {}

          public void onResponse(StreamingRecognizeResponse response) {
            responses.add(response);
          }

          public void onComplete() {
            for (StreamingRecognizeResponse response : responses) {
              StreamingRecognitionResult result = response.getResultsList().get(0);
              SpeechRecognitionAlternative alternative = result.getAlternativesList().get(0);
              System.out.printf("Transcript : %s\n", alternative.getTranscript());
            }
          }

          public void onError(Throwable t) {
            System.out.println(t);
          }
        };

    ClientStream<StreamingRecognizeRequest> clientStream =
        client.streamingRecognizeCallable().splitCall(responseObserver);

    RecognitionConfig recognitionConfig =
        RecognitionConfig.newBuilder()
            .setEncoding(RecognitionConfig.AudioEncoding.LINEAR16)
            .setLanguageCode("en-US")
            .setSampleRateHertz(16000)
            .build();
    StreamingRecognitionConfig streamingRecognitionConfig =
        StreamingRecognitionConfig.newBuilder().setConfig(recognitionConfig).build();

    StreamingRecognizeRequest request =
        StreamingRecognizeRequest.newBuilder()
            .setStreamingConfig(streamingRecognitionConfig)
            .build(); // The first request in a streaming call has to be a config

    clientStream.send(request);
    // SampleRate:16000Hz, SampleSizeInBits: 16, Number of channels: 1, Signed: true,
    // bigEndian: false
    AudioFormat audioFormat = new AudioFormat(16000, 16, 1, true, false);
    DataLine.Info targetInfo =
        new Info(
            TargetDataLine.class,
            audioFormat); // Set the system information to read from the microphone audio stream

    if (!AudioSystem.isLineSupported(targetInfo)) {
      System.out.println("Microphone not supported");
      System.exit(0);
    }
    // Target data line captures the audio stream the microphone produces.
    TargetDataLine targetDataLine = (TargetDataLine) AudioSystem.getLine(targetInfo);
    targetDataLine.open(audioFormat);
    targetDataLine.start();
    System.out.println("Start speaking");
    long startTime = System.currentTimeMillis();
    // Audio Input Stream
    AudioInputStream audio = new AudioInputStream(targetDataLine);
    while (true) {
      long estimatedTime = System.currentTimeMillis() - startTime;
      byte[] data = new byte[6400];
      audio.read(data);
      if (estimatedTime > 60000) { // 60 seconds
        System.out.println("Stop speaking.");
        targetDataLine.stop();
        targetDataLine.close();
        break;
      }
      request =
          StreamingRecognizeRequest.newBuilder()
              .setAudioContent(ByteString.copyFrom(data))
              .build();
      clientStream.send(request);
    }
  } catch (Exception e) {
    System.out.println(e);
  }
  responseObserver.onComplete();
}

Node.js

This samples requires you to install SoX and it must be available in your $PATH.

  • For Mac OS: brew install sox.
  • For most Linux distributions: sudo apt-get install sox libsox-fmt-all.
  • For Windows: Download the binaries.

For more on installing and creating a Speech-to-Text client, refer to Speech-to-Text Client Libraries.

const recorder = require('node-record-lpcm16');

// Imports the Google Cloud client library
const speech = require('@google-cloud/speech');

// Creates a client
const client = new speech.SpeechClient();

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const encoding = 'Encoding of the audio file, e.g. LINEAR16';
// const sampleRateHertz = 16000;
// const languageCode = 'BCP-47 language code, e.g. en-US';

const request = {
  config: {
    encoding: encoding,
    sampleRateHertz: sampleRateHertz,
    languageCode: languageCode,
  },
  interimResults: false, // If you want interim results, set this to true
};

// Create a recognize stream
const recognizeStream = client
  .streamingRecognize(request)
  .on('error', console.error)
  .on('data', data =>
    process.stdout.write(
      data.results[0] && data.results[0].alternatives[0]
        ? `Transcription: ${data.results[0].alternatives[0].transcript}\n`
        : '\n\nReached transcription time limit, press Ctrl+C\n'
    )
  );

// Start recording and send the microphone input to the Speech API.
// Ensure SoX is installed, see https://www.npmjs.com/package/node-record-lpcm16#dependencies
recorder
  .record({
    sampleRateHertz: sampleRateHertz,
    threshold: 0,
    // Other options, see https://www.npmjs.com/package/node-record-lpcm16#options
    verbose: false,
    recordProgram: 'rec', // Try also "arecord" or "sox"
    silence: '10.0',
  })
  .stream()
  .on('error', console.error)
  .pipe(recognizeStream);

console.log('Listening, press Ctrl+C to stop.');

What's next