Java SynthesizeSpeechRequest类代码示例

import com.amazonaws.services.polly.model.SynthesizeSpeechRequest; //导入依赖的package包/类
public byte[] cacheFile(String toSpeak, OutputFormat format) throws IOException {

    byte[] mp3File = null;
    // cache it begin -----
    String localFileName = getLocalFileName(this, toSpeak, "mp3");
    // String filename = AudioFile.globalFileCacheDir + File.separator +
    // localFileName;
    if (!audioFile.cacheContains(localFileName)) {
      log.info("retrieving speech from Amazon - {}", localFileName);
      AmazonPollyClient polly = getPolly();
      SynthesizeSpeechRequest synthReq = new SynthesizeSpeechRequest().withText(toSpeak).withVoiceId(awsVoice.getId()).withOutputFormat(format);
      SynthesizeSpeechResult synthRes = polly.synthesizeSpeech(synthReq);
      InputStream data = synthRes.getAudioStream();
      mp3File = FileIO.toByteArray(data);
      audioFile.cache(localFileName, mp3File, toSpeak);
    } else {
      log.info("using local cached file");
      mp3File = FileIO.toByteArray(new File(AudioFile.globalFileCacheDir + File.separator + getLocalFileName(this, toSpeak, "mp3")));
    }

    // invoke("publishStartSpeaking", toSpeak);
    // audioFile.playBlocking(filename);
    // invoke("publishEndSpeaking", toSpeak);
    // log.info("Finished waiting for completion.");
    return mp3File; 
  }
 

import com.amazonaws.services.polly.model.SynthesizeSpeechRequest; //导入依赖的package包/类
@Override
public void handle(Context ctx) throws Exception {
    String voiceId = ctx.getRequest().getQueryParams().get("voiceId");
    String text = ctx.getRequest().getQueryParams().get("text");
    String outputFormat = ctx.getRequest().getQueryParams().get("outputFormat");

    SynthesizeSpeechRequest ssRequest = new SynthesizeSpeechRequest();
    ssRequest.setVoiceId(voiceId);
    ssRequest.setOutputFormat(outputFormat);
    ssRequest.setText(text);

    SynthesizeSpeechResult result = polly.synthesizeSpeech(ssRequest);

    ctx.getResponse().contentType(result.getContentType());
    ctx.getResponse().sendStream(s -> s.onSubscribe(new Subscription() {
        @Override
        public void request(long n) {
            try {
                byte[] data      = new byte[1024];
                int bytesRead = result.getAudioStream().read(data);

                while(bytesRead != -1) {
                    s.onNext(Unpooled.wrappedBuffer(data));
                    bytesRead = result.getAudioStream().read(data);
                }
            } catch (IOException e) {
                ctx.getResponse().status(500);
                ctx.getResponse().send();
            } finally {
                s.onComplete();
            }
        }

        @Override
        public void cancel() {

        }
    }));
}
 

import com.amazonaws.services.polly.model.SynthesizeSpeechRequest; //导入依赖的package包/类
/**
 * Returns text-to-speech of a translation of a given text. Before translating the text,
 * requesting speech from AWS Polly and storing the resulting MP3 to S3 this method looks
 * up previous translation of the same text. Once found it will avoid doing the aforementioned
 * roundtrip but rather will use the data of the previous translation.
 * @param text text to translate and convert to speech
 * @return text to speech information
 * @throws AlexaStateException error reading or writing state to Dynamo dictionary
 */
public Optional<TextToSpeech> textToSpeech(final String text) throws AlexaStateException {
    // remove invalid prefixes that accidently made it into the slots
    final String textToTranslate = prefixesToRemove.stream()
            .filter(prefix -> StringUtils.startsWithIgnoreCase(text, prefix))
            .findFirst()
            .map(prefix -> text.replaceFirst(prefix, ""))
            // if none of these prefixes exist in the text, keep the text as is
            .orElse(text);

    // look up previous translation in dictionary
    Optional<TextToSpeech> tts = dynamoStateHandler.readModel(TextToSpeech.class, getDictionaryId(textToTranslate));
    // if there was a previous tts for this text return immediately (exception for the roundtrip-phrase used by the test-client)
    if (tts.isPresent() && !StringUtils.equalsIgnoreCase(textToTranslate, SkillConfig.getAlwaysRoundTripPhrase())) {
        // set handler to session to avoid writing back to dynamo (nothing changed)
        tts.get().setHandler(sessionStateHandler);
        return tts;
    }

    // translate term by leveraging a Translator implementation provided by the factory
    final Optional<String> translated = translator.translate(textToTranslate, language);

    if (translated.isPresent()) {
        // without a voiceId there's not chance to fulfill the translation request
        Validate.notBlank(voiceId, "No voiceId is associated with given language.");
        // form the SSML by embedding the translated text
        final String ssml = String.format("<speak><amazon:effect name='drc'><prosody rate='-15%%' volume='x-loud'>%1$s</prosody></amazon:effect><break time='250ms' /></speak>", translated.get());
        // build a Polly request to get speech with desired voice and SSML
        final SynthesizeSpeechRequest synthRequest = new SynthesizeSpeechRequest()
                .withText(ssml)
                .withOutputFormat(OutputFormat.Mp3)
                .withVoiceId(voiceId)
                .withTextType(TextType.Ssml)
                .withSampleRate("22050");
        // fire request to Polly
        final SynthesizeSpeechResult synthResult = awsPolly.synthesizeSpeech(synthRequest);

        try {
            // store audio stream of Polly to S3 as an MP3 file
            final PutObjectRequest s3Put = new PutObjectRequest(SkillConfig.getS3BucketName(), getMp3Path(textToTranslate), synthResult.getAudioStream(), new ObjectMetadata())
                    .withCannedAcl(CannedAccessControlList.PublicRead);
            awsS3.putObject(s3Put);
            // as long as Polly output does not comply with Alexa MP3 format restriction we need to convert the MP3
            if (!SkillConfig.shouldSkipMp3Conversion()) {
                // call the REST service that encapsualtes the FFMPEG conversion on a server
                final String mp3ConvertedUrl = Mp3Converter.convertMp3(getMp3Path(textToTranslate));
                // validate this service returned a url (equal to success)
                Validate.notBlank(mp3ConvertedUrl, "Conversion service did not return proper return value");
            }
            // build the TTS object with all the information needed to return output speech
            return Optional.of(getTTS(textToTranslate, translated.get()));
        } catch (final IOException | URISyntaxException e) {
            log.error("Error while generating mp3. " + e.getMessage());
        }
    }
    return Optional.empty();
}