ストリーミング入力の音声を文字に変換する

このセクションでは、マイクからの入力などのストリーミング音声をテキストに変換する方法について説明します。

ストリーミング音声認識では、音声を Speech-to-Text にストリーミングし、音声を処理しながらリアルタイムでストリーム音声認識の結果を受信できます。ストリーミング音声認識リクエストについては、音声の制限もご覧ください。ストリーミング音声認識は、gRPC 経由でのみ利用できます。

ローカル ファイルでストリーミング音声認識を実行する

ローカル音声ファイルに対して、ストリーミング音声認識を実行する例を次に示します。API に送信されるすべてのストリーミング リクエストには 10 MB の上限があります。この上限は、最初の StreamingRecognize リクエストと、ストリーム内の各メッセージのサイズの両方に適用されます。この上限を超えると、エラーがスローされます。

Go

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

	speech "cloud.google.com/go/speech/apiv1"
	"cloud.google.com/go/speech/apiv1/speechpb"
)

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);

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))

その他の言語

C#: クライアント ライブラリ ページの C# の設定手順を行ってから、.NET の Speech-to-Text のリファレンス ドキュメントをご覧ください。

PHP: クライアント ライブラリ ページの PHP の設定手順を行ってから、PHP の Speech-to-Text のリファレンス ドキュメントをご覧ください。

Ruby: クライアント ライブラリ ページの Ruby の設定手順を行ってから、Ruby の Speech-to-Text のリファレンス ドキュメントをご覧ください。

ローカルの音声ファイルを Speech-to-Text API にストリーミングすることは可能ですが、同期または非同期の音声認識を行ってバッチモードの結果を取得することをおすすめします。

音声ストリームでストリーミング音声認識を実行する

Speech-to-Text では、リアルタイムのストリーミング音声の認識も行うことができます。

マイクから受信した音声ストリームに対して、ストリーミング音声認識を実行する例を次に示します。

Go

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

	speech "cloud.google.com/go/speech/apiv1"
	"cloud.google.com/go/speech/apiv1/speechpb"
)

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)
		}
	}
}

Python

from __future__ import division

import re
import sys

from google.cloud import speech

import pyaudio
from six.moves import queue

# Audio recording parameters
RATE = 16000
CHUNK = int(RATE / 10)  # 100ms

class MicrophoneStream(object):
    """Opens a recording stream as a generator yielding the audio chunks."""

    def __init__(self, rate, chunk):
        self._rate = rate
        self._chunk = chunk

        # Create a thread-safe buffer of audio data
        self._buff = queue.Queue()
        self.closed = True

    def __enter__(self):
        self._audio_interface = pyaudio.PyAudio()
        self._audio_stream = self._audio_interface.open(
            format=pyaudio.paInt16,
            # The API currently only supports 1-channel (mono) audio
            # https://goo.gl/z757pE
            channels=1,
            rate=self._rate,
            input=True,
            frames_per_buffer=self._chunk,
            # Run the audio stream asynchronously to fill the buffer object.
            # This is necessary so that the input device's buffer doesn't
            # overflow while the calling thread makes network requests, etc.
            stream_callback=self._fill_buffer,
        )

        self.closed = False

        return self

    def __exit__(self, type, value, traceback):
        self._audio_stream.stop_stream()
        self._audio_stream.close()
        self.closed = True
        # Signal the generator to terminate so that the client's
        # streaming_recognize method will not block the process termination.
        self._buff.put(None)
        self._audio_interface.terminate()

    def _fill_buffer(self, in_data, frame_count, time_info, status_flags):
        """Continuously collect data from the audio stream, into the buffer."""
        self._buff.put(in_data)
        return None, pyaudio.paContinue

    def generator(self):
        while not self.closed:
            # Use a blocking get() to ensure there's at least one chunk of
            # data, and stop iteration if the chunk is None, indicating the
            # end of the audio stream.
            chunk = self._buff.get()
            if chunk is None:
                return
            data = [chunk]

            # Now consume whatever other data's still buffered.
            while True:
                try:
                    chunk = self._buff.get(block=False)
                    if chunk is None:
                        return
                    data.append(chunk)
                except queue.Empty:
                    break

            yield b"".join(data)

def listen_print_loop(responses):
    """Iterates through server responses and prints them.

    The responses passed is a generator that will block until a response
    is provided by the server.

    Each response may contain multiple results, and each result may contain
    multiple alternatives; for details, see https://goo.gl/tjCPAU.  Here we
    print only the transcription for the top alternative of the top result.

    In this case, responses are provided for interim results as well. If the
    response is an interim one, print a line feed at the end of it, to allow
    the next result to overwrite it, until the response is a final one. For the
    final one, print a newline to preserve the finalized transcription.
    """
    num_chars_printed = 0
    for response in responses:
        if not response.results:
            continue

        # The `results` list is consecutive. For streaming, we only care about
        # the first result being considered, since once it's `is_final`, it
        # moves on to considering the next utterance.
        result = response.results[0]
        if not result.alternatives:
            continue

        # Display the transcription of the top alternative.
        transcript = result.alternatives[0].transcript

        # Display interim results, but with a carriage return at the end of the
        # line, so subsequent lines will overwrite them.
        #
        # If the previous result was longer than this one, we need to print
        # some extra spaces to overwrite the previous result
        overwrite_chars = " " * (num_chars_printed - len(transcript))

        if not result.is_final:
            sys.stdout.write(transcript + overwrite_chars + "\r")
            sys.stdout.flush()

            num_chars_printed = len(transcript)

        else:
            print(transcript + overwrite_chars)

            # Exit recognition if any of the transcribed phrases could be
            # one of our keywords.
            if re.search(r"\b(exit|quit)\b", transcript, re.I):
                print("Exiting..")
                break

            num_chars_printed = 0

def main():
    # See http://g.co/cloud/speech/docs/languages
    # for a list of supported languages.
    language_code = "en-US"  # a BCP-47 language tag

    client = speech.SpeechClient()
    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=RATE,
        language_code=language_code,
    )

    streaming_config = speech.StreamingRecognitionConfig(
        config=config, interim_results=True
    )

    with MicrophoneStream(RATE, CHUNK) as stream:
        audio_generator = stream.generator()
        requests = (
            speech.StreamingRecognizeRequest(audio_content=content)
            for content in audio_generator
        )

        responses = client.streaming_recognize(streaming_config, requests)

        # Now, put the transcription responses to use.
        listen_print_loop(responses)

if __name__ == "__main__":
    main()

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

このサンプルを使用するには、SoX をインストールして、$PATH で使用できるようにする必要があります。

  • macOS の場合: brew install sox
  • ほとんどの Linux ディストリビューションの場合: sudo apt-get install sox libsox-fmt-all
  • Windows の場合はバイナリをダウンロード

Speech-to-Text クライアントのインストールと作成の詳細については、Speech-to-Text クライアント ライブラリをご覧ください。

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.');

その他の言語

C#: クライアント ライブラリ ページの C# の設定手順を行ってから、.NET の Speech-to-Text のリファレンス ドキュメントをご覧ください。

PHP: クライアント ライブラリ ページの PHP の設定手順を行ってから、PHP の Speech-to-Text のリファレンス ドキュメントをご覧ください。

Ruby: クライアント ライブラリ ページの Ruby の設定手順を行ってから、Ruby の Speech-to-Text のリファレンス ドキュメントをご覧ください。

次のステップ

使ってみる

Google Cloud を初めて使用する場合は、アカウントを作成して、実際のシナリオでの Speech-to-Text のパフォーマンスを評価してください。新規のお客様には、ワークロードの実行、テスト、デプロイができる無料クレジット $300 分を差し上げます。

Speech-to-Text の無料トライアル