git.basschouten.com Git - openhab-addons.git/blob

   1 /**
   2  * Copyright (c) 2010-2023 Contributors to the openHAB project
   3  *
   4  * See the NOTICE file(s) distributed with this work for additional
   5  * information.
   6  *
   7  * This program and the accompanying materials are made available under the
   8  * terms of the Eclipse Public License 2.0 which is available at
   9  * http://www.eclipse.org/legal/epl-2.0
  10  *
  11  * SPDX-License-Identifier: EPL-2.0
  12  */
  13 package org.openhab.voice.voicerss.internal.cloudapi;
  14
  15 import java.io.IOException;
  16 import java.io.InputStream;
  17 import java.net.HttpURLConnection;
  18 import java.net.URL;
  19 import java.net.URLConnection;
  20 import java.net.URLEncoder;
  21 import java.nio.charset.StandardCharsets;
  22 import java.util.HashMap;
  23 import java.util.HashSet;
  24 import java.util.List;
  25 import java.util.Locale;
  26 import java.util.Map;
  27 import java.util.Map.Entry;
  28 import java.util.Set;
  29
  30 import org.eclipse.jdt.annotation.NonNullByDefault;
  31 import org.openhab.core.audio.AudioFormat;
  32 import org.openhab.voice.voicerss.internal.VoiceRSSRawAudioStream;
  33 import org.slf4j.Logger;
  34 import org.slf4j.LoggerFactory;
  35
  36 /**
  37  * This class implements the Cloud service from VoiceRSS. For more information,
  38  * see API documentation at http://www.voicerss.org/api .
  39  *
  40  * Current state of implementation:
  41  * <ul>
  42  * <li>All API languages supported</li>
  43  * <li>Only default voice supported with good audio quality</li>
  44  * <li>MP3, OGG, AAC and WAV audio formats supported</li>
  45  * <li>It uses HTTP and not HTTPS (for performance reasons)</li>
  46  * </ul>
  47  *
  48  * @author Jochen Hiller - Initial contribution
  49  * @author Laurent Garnier - add support for all API languages
  50  * @author Laurent Garnier - add support for OGG and AAC audio formats
  51  * @author Andreas Brenk - add support for WAV audio format
  52  */
  53 @NonNullByDefault
  54 public class VoiceRSSCloudImpl implements VoiceRSSCloudAPI {
  55
  56     public static final String DEFAULT_VOICE = "default";
  57
  58     public static final String API_URL = "https://api.voicerss.org/?key=%s&hl=%s&c=%s&f=%s&src=%s";
  59     public static final String API_URL_WITH_VOICE = API_URL + "&v=%s";
  60
  61     private static final Set<String> SUPPORTED_AUDIO_CODECS = Set.of("MP3", "OGG", "AAC", "WAV", "CAF");
  62
  63     private static final Set<Locale> SUPPORTED_LOCALES = new HashSet<>();
  64     static {
  65         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ar-eg"));
  66         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ar-sa"));
  67         SUPPORTED_LOCALES.add(Locale.forLanguageTag("bg-bg"));
  68         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ca-es"));
  69         SUPPORTED_LOCALES.add(Locale.forLanguageTag("cs-cz"));
  70         SUPPORTED_LOCALES.add(Locale.forLanguageTag("da-dk"));
  71         SUPPORTED_LOCALES.add(Locale.forLanguageTag("de-at"));
  72         SUPPORTED_LOCALES.add(Locale.forLanguageTag("de-de"));
  73         SUPPORTED_LOCALES.add(Locale.forLanguageTag("de-ch"));
  74         SUPPORTED_LOCALES.add(Locale.forLanguageTag("el-gr"));
  75         SUPPORTED_LOCALES.add(Locale.forLanguageTag("en-au"));
  76         SUPPORTED_LOCALES.add(Locale.forLanguageTag("en-ca"));
  77         SUPPORTED_LOCALES.add(Locale.forLanguageTag("en-gb"));
  78         SUPPORTED_LOCALES.add(Locale.forLanguageTag("en-ie"));
  79         SUPPORTED_LOCALES.add(Locale.forLanguageTag("en-in"));
  80         SUPPORTED_LOCALES.add(Locale.forLanguageTag("en-us"));
  81         SUPPORTED_LOCALES.add(Locale.forLanguageTag("es-es"));
  82         SUPPORTED_LOCALES.add(Locale.forLanguageTag("es-mx"));
  83         SUPPORTED_LOCALES.add(Locale.forLanguageTag("fi-fi"));
  84         SUPPORTED_LOCALES.add(Locale.forLanguageTag("fr-ca"));
  85         SUPPORTED_LOCALES.add(Locale.forLanguageTag("fr-fr"));
  86         SUPPORTED_LOCALES.add(Locale.forLanguageTag("fr-ch"));
  87         SUPPORTED_LOCALES.add(Locale.forLanguageTag("he-il"));
  88         SUPPORTED_LOCALES.add(Locale.forLanguageTag("hi-in"));
  89         SUPPORTED_LOCALES.add(Locale.forLanguageTag("hr-hr"));
  90         SUPPORTED_LOCALES.add(Locale.forLanguageTag("hu-hu"));
  91         SUPPORTED_LOCALES.add(Locale.forLanguageTag("id-id"));
  92         SUPPORTED_LOCALES.add(Locale.forLanguageTag("it-it"));
  93         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ja-jp"));
  94         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ko-kr"));
  95         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ms-my"));
  96         SUPPORTED_LOCALES.add(Locale.forLanguageTag("nb-no"));
  97         SUPPORTED_LOCALES.add(Locale.forLanguageTag("nl-be"));
  98         SUPPORTED_LOCALES.add(Locale.forLanguageTag("nl-nl"));
  99         SUPPORTED_LOCALES.add(Locale.forLanguageTag("pl-pl"));
 100         SUPPORTED_LOCALES.add(Locale.forLanguageTag("pt-br"));
 101         SUPPORTED_LOCALES.add(Locale.forLanguageTag("pt-pt"));
 102         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ro-ro"));
 103         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ru-ru"));
 104         SUPPORTED_LOCALES.add(Locale.forLanguageTag("sk-sk"));
 105         SUPPORTED_LOCALES.add(Locale.forLanguageTag("sl-si"));
 106         SUPPORTED_LOCALES.add(Locale.forLanguageTag("sv-se"));
 107         SUPPORTED_LOCALES.add(Locale.forLanguageTag("ta-in"));
 108         SUPPORTED_LOCALES.add(Locale.forLanguageTag("th-th"));
 109         SUPPORTED_LOCALES.add(Locale.forLanguageTag("tr-tr"));
 110         SUPPORTED_LOCALES.add(Locale.forLanguageTag("vi-vn"));
 111         SUPPORTED_LOCALES.add(Locale.forLanguageTag("zh-cn"));
 112         SUPPORTED_LOCALES.add(Locale.forLanguageTag("zh-hk"));
 113         SUPPORTED_LOCALES.add(Locale.forLanguageTag("zh-tw"));
 114     }
 115
 116     private static final Map<String, Set<String>> SUPPORTED_VOICES = new HashMap<>();
 117     static {
 118         SUPPORTED_VOICES.put("ar-eg", Set.of("Oda"));
 119         SUPPORTED_VOICES.put("ar-sa", Set.of("Salim"));
 120         SUPPORTED_VOICES.put("bg-bg", Set.of("Dimo"));
 121         SUPPORTED_VOICES.put("ca-es", Set.of("Rut"));
 122         SUPPORTED_VOICES.put("cs-cz", Set.of("Josef"));
 123         SUPPORTED_VOICES.put("da-dk", Set.of("Freja"));
 124         SUPPORTED_VOICES.put("de-at", Set.of("Lukas"));
 125         SUPPORTED_VOICES.put("de-de", Set.of("Hanna", "Lina", "Jonas"));
 126         SUPPORTED_VOICES.put("de-ch", Set.of("Tim"));
 127         SUPPORTED_VOICES.put("el-gr", Set.of("Neo"));
 128         SUPPORTED_VOICES.put("en-au", Set.of("Zoe", "Isla", "Evie", "Jack"));
 129         SUPPORTED_VOICES.put("en-ca", Set.of("Rose", "Clara", "Emma", "Mason"));
 130         SUPPORTED_VOICES.put("en-gb", Set.of("Alice", "Nancy", "Lily", "Harry"));
 131         SUPPORTED_VOICES.put("en-ie", Set.of("Oran"));
 132         SUPPORTED_VOICES.put("en-in", Set.of("Eka", "Jai", "Ajit"));
 133         SUPPORTED_VOICES.put("en-us", Set.of("Linda", "Amy", "Mary", "John", "Mike"));
 134         SUPPORTED_VOICES.put("es-es", Set.of("Camila", "Sofia", "Luna", "Diego"));
 135         SUPPORTED_VOICES.put("es-mx", Set.of("Juana", "Silvia", "Teresa", "Jose"));
 136         SUPPORTED_VOICES.put("fi-fi", Set.of("Aada"));
 137         SUPPORTED_VOICES.put("fr-ca", Set.of("Emile", "Olivia", "Logan", "Felix"));
 138         SUPPORTED_VOICES.put("fr-fr", Set.of("Bette", "Iva", "Zola", "Axel"));
 139         SUPPORTED_VOICES.put("fr-ch", Set.of("Theo"));
 140         SUPPORTED_VOICES.put("he-il", Set.of("Rami"));
 141         SUPPORTED_VOICES.put("hi-in", Set.of("Puja", "Kabir"));
 142         SUPPORTED_VOICES.put("hr-hr", Set.of("Nikola"));
 143         SUPPORTED_VOICES.put("hu-hu", Set.of("Mate"));
 144         SUPPORTED_VOICES.put("id-id", Set.of("Intan"));
 145         SUPPORTED_VOICES.put("it-it", Set.of("Bria", "Mia", "Pietro"));
 146         SUPPORTED_VOICES.put("ja-jp", Set.of("Hina", "Airi", "Fumi", "Akira"));
 147         SUPPORTED_VOICES.put("ko-kr", Set.of("Nari"));
 148         SUPPORTED_VOICES.put("ms-my", Set.of("Aqil"));
 149         SUPPORTED_VOICES.put("nb-no", Set.of("Marte", "Erik"));
 150         SUPPORTED_VOICES.put("nl-be", Set.of("Daan"));
 151         SUPPORTED_VOICES.put("nl-nl", Set.of("Lotte", "Bram"));
 152         SUPPORTED_VOICES.put("pl-pl", Set.of("Julia", "Jan"));
 153         SUPPORTED_VOICES.put("pt-br", Set.of("Marcia", "Ligia", "Yara", "Dinis"));
 154         SUPPORTED_VOICES.put("pt-pt", Set.of("Leonor"));
 155         SUPPORTED_VOICES.put("ro-ro", Set.of("Doru"));
 156         SUPPORTED_VOICES.put("ru-ru", Set.of("Olga", "Marina", "Peter"));
 157         SUPPORTED_VOICES.put("sk-sk", Set.of("Beda"));
 158         SUPPORTED_VOICES.put("sl-si", Set.of("Vid"));
 159         SUPPORTED_VOICES.put("sv-se", Set.of("Molly", "Hugo"));
 160         SUPPORTED_VOICES.put("ta-in", Set.of("Sai"));
 161         SUPPORTED_VOICES.put("th-th", Set.of("Ukrit"));
 162         SUPPORTED_VOICES.put("tr-tr", Set.of("Omer"));
 163         SUPPORTED_VOICES.put("vi-vn", Set.of("Chi"));
 164         SUPPORTED_VOICES.put("zh-cn", Set.of("Luli", "Shu", "Chow", "Wang"));
 165         SUPPORTED_VOICES.put("zh-hk", Set.of("Jia", "Xia", "Chen"));
 166         SUPPORTED_VOICES.put("zh-tw", Set.of("Akemi", "Lin", "Lee"));
 167     }
 168
 169     protected boolean logging;
 170
 171     public VoiceRSSCloudImpl(boolean logging) {
 172         this.logging = logging;
 173     }
 174
 175     @Override
 176     public Set<String> getAvailableAudioCodecs() {
 177         return SUPPORTED_AUDIO_CODECS;
 178     }
 179
 180     @Override
 181     public Set<Locale> getAvailableLocales() {
 182         return SUPPORTED_LOCALES;
 183     }
 184
 185     @Override
 186     public Set<String> getAvailableVoices() {
 187         // different locales support different voices, so let's list all here in one big set when no locale is provided
 188         Set<String> allvoxes = new HashSet<>();
 189         allvoxes.add(DEFAULT_VOICE);
 190         for (Set<String> langvoxes : SUPPORTED_VOICES.values()) {
 191             for (String langvox : langvoxes) {
 192                 allvoxes.add(langvox);
 193             }
 194         }
 195         return allvoxes;
 196     }
 197
 198     @Override
 199     public Set<String> getAvailableVoices(Locale locale) {
 200         Set<String> allvoxes = new HashSet<>();
 201         allvoxes.add(DEFAULT_VOICE);
 202         // all maps must be defined with key in lowercase
 203         String langtag = locale.toLanguageTag().toLowerCase();
 204         if (SUPPORTED_VOICES.containsKey(langtag)) {
 205             for (String langvox : SUPPORTED_VOICES.get(langtag)) {
 206                 allvoxes.add(langvox);
 207             }
 208         }
 209         return allvoxes;
 210     }
 211
 212     /*
 213      * This method will return an input stream to an audio stream for the given
 214      * parameters.
 215      *
 216      * It will do that using a plain URL connection to avoid any external
 217      * dependencies.
 218      */
 219     @Override
 220     public VoiceRSSRawAudioStream getTextToSpeech(String apiKey, String text, String locale, String voice,
 221             String audioCodec, String audioFormat) throws IOException {
 222         String url = createURL(apiKey, text, locale, voice, audioCodec, audioFormat);
 223         if (logging) {
 224             LoggerFactory.getLogger(VoiceRSSCloudImpl.class).debug("Call {}", url.replace(apiKey, "***"));
 225         }
 226         URLConnection connection = new URL(url).openConnection();
 227
 228         // we will check return codes. The service will ALWAYS return a HTTP
 229         // 200, but for error messages, it will return a text/plain format and
 230         // the error message in body
 231         int status = ((HttpURLConnection) connection).getResponseCode();
 232         if (HttpURLConnection.HTTP_OK != status) {
 233             if (logging) {
 234                 LoggerFactory.getLogger(VoiceRSSCloudImpl.class).warn("Call {} returned HTTP {}",
 235                         url.replace(apiKey, "***"), status);
 236             }
 237             throw new IOException("Could not read from service: HTTP code " + status);
 238         }
 239         if (logging) {
 240             Logger logger = LoggerFactory.getLogger(VoiceRSSCloudImpl.class);
 241             if (logger.isTraceEnabled()) {
 242                 for (Entry<String, List<String>> header : connection.getHeaderFields().entrySet()) {
 243                     logger.trace("Response.header: {}={}", header.getKey(), header.getValue());
 244                 }
 245             }
 246         }
 247         String contentType = connection.getHeaderField("Content-Type");
 248         InputStream is = connection.getInputStream();
 249         // check if content type is text/plain, then we have an error
 250         if (contentType.contains("text/plain")) {
 251             byte[] bytes = new byte[256];
 252             is.read(bytes, 0, 256);
 253             // close before throwing an exception
 254             try {
 255                 is.close();
 256             } catch (IOException ex) {
 257                 if (logging) {
 258                     LoggerFactory.getLogger(VoiceRSSCloudImpl.class).debug("Failed to close inputstream", ex);
 259                 }
 260             }
 261             throw new IOException(
 262                     "Could not read audio content, service returned an error: " + new String(bytes, "UTF-8"));
 263         } else {
 264             // Set any audio format
 265             return new VoiceRSSRawAudioStream(is, AudioFormat.MP3, connection.getContentLengthLong());
 266         }
 267     }
 268
 269     // internal
 270
 271     /**
 272      * This method will create the URL for the cloud service. The text will be
 273      * URI encoded as it is part of the URL.
 274      *
 275      * It is in package scope to be accessed by tests.
 276      */
 277     private String createURL(String apiKey, String text, String locale, String voice, String audioCodec,
 278             String audioFormat) {
 279         String encodedMsg = URLEncoder.encode(text, StandardCharsets.UTF_8);
 280         String url;
 281         if (!DEFAULT_VOICE.equals(voice)) {
 282             url = String.format(API_URL_WITH_VOICE, apiKey, locale, audioCodec, audioFormat, encodedMsg, voice);
 283         } else {
 284             url = String.format(API_URL, apiKey, locale, audioCodec, audioFormat, encodedMsg);
 285         }
 286         return url;
 287     }
 288 }