echogarden 0.0.1 → 0.1.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (295) hide show
  1. package/README.md +59 -1
  2. package/data/lexicons/heteronyms.json +674 -0
  3. package/data/schemas/options.json +1057 -0
  4. package/data/tables/lcid-table.json +7427 -0
  5. package/dist/alignment/DTWMfccSequenceAlignment.d.ts +2 -0
  6. package/dist/alignment/DTWMfccSequenceAlignment.js +26 -0
  7. package/dist/alignment/DTWMfccSequenceAlignment.js.map +1 -0
  8. package/dist/alignment/DTWSequenceAlignment.d.ts +5 -0
  9. package/dist/alignment/DTWSequenceAlignment.js +99 -0
  10. package/dist/alignment/DTWSequenceAlignment.js.map +1 -0
  11. package/dist/alignment/DTWSequenceAlignmentWindowed.d.ts +5 -0
  12. package/dist/alignment/DTWSequenceAlignmentWindowed.js +161 -0
  13. package/dist/alignment/DTWSequenceAlignmentWindowed.js.map +1 -0
  14. package/dist/alignment/LevenshteinSequenceAlignment.d.ts +5 -0
  15. package/dist/alignment/LevenshteinSequenceAlignment.js +102 -0
  16. package/dist/alignment/LevenshteinSequenceAlignment.js.map +1 -0
  17. package/dist/alignment/SpeechAlignment.d.ts +27 -0
  18. package/dist/alignment/SpeechAlignment.js +266 -0
  19. package/dist/alignment/SpeechAlignment.js.map +1 -0
  20. package/dist/api/API.d.ts +8 -0
  21. package/dist/api/API.js +10 -0
  22. package/dist/api/API.js.map +1 -0
  23. package/dist/api/APIOptions.d.ts +12 -0
  24. package/dist/api/APIOptions.js +2 -0
  25. package/dist/api/APIOptions.js.map +1 -0
  26. package/dist/api/Alignment.d.ts +32 -0
  27. package/dist/api/Alignment.js +134 -0
  28. package/dist/api/Alignment.js.map +1 -0
  29. package/dist/api/Denoising.d.ts +14 -0
  30. package/dist/api/Denoising.js +66 -0
  31. package/dist/api/Denoising.js.map +1 -0
  32. package/dist/api/Globals.d.ts +1 -0
  33. package/dist/api/Globals.js +2 -0
  34. package/dist/api/Globals.js.map +1 -0
  35. package/dist/api/LanguageDetection.d.ts +49 -0
  36. package/dist/api/LanguageDetection.js +118 -0
  37. package/dist/api/LanguageDetection.js.map +1 -0
  38. package/dist/api/Recognition.d.ts +49 -0
  39. package/dist/api/Recognition.js +162 -0
  40. package/dist/api/Recognition.js.map +1 -0
  41. package/dist/api/Synthesis.d.ts +125 -0
  42. package/dist/api/Synthesis.js +892 -0
  43. package/dist/api/Synthesis.js.map +1 -0
  44. package/dist/api/Translation.d.ts +25 -0
  45. package/dist/api/Translation.js +69 -0
  46. package/dist/api/Translation.js.map +1 -0
  47. package/dist/api/Vad.d.ts +25 -0
  48. package/dist/api/Vad.js +88 -0
  49. package/dist/api/Vad.js.map +1 -0
  50. package/dist/audio/AudioBufferConversion.d.ts +15 -0
  51. package/dist/audio/AudioBufferConversion.js +228 -0
  52. package/dist/audio/AudioBufferConversion.js.map +1 -0
  53. package/dist/audio/AudioPlayer.d.ts +9 -0
  54. package/dist/audio/AudioPlayer.js +232 -0
  55. package/dist/audio/AudioPlayer.js.map +1 -0
  56. package/dist/audio/AudioRecorder.d.ts +3 -0
  57. package/dist/audio/AudioRecorder.js +68 -0
  58. package/dist/audio/AudioRecorder.js.map +1 -0
  59. package/dist/audio/AudioUtilities.d.ts +48 -0
  60. package/dist/audio/AudioUtilities.js +209 -0
  61. package/dist/audio/AudioUtilities.js.map +1 -0
  62. package/dist/audio/SoxPath.d.ts +1 -0
  63. package/dist/audio/SoxPath.js +19 -0
  64. package/dist/audio/SoxPath.js.map +1 -0
  65. package/dist/cli/CLI.d.ts +8 -0
  66. package/dist/cli/CLI.js +860 -0
  67. package/dist/cli/CLI.js.map +1 -0
  68. package/dist/cli/CLIConfigFile.d.ts +3 -0
  69. package/dist/cli/CLIConfigFile.js +62 -0
  70. package/dist/cli/CLIConfigFile.js.map +1 -0
  71. package/dist/cli/CLILauncher.d.ts +2 -0
  72. package/dist/cli/CLILauncher.js +22 -0
  73. package/dist/cli/CLILauncher.js.map +1 -0
  74. package/dist/cli/CLIOptionsSchema.d.ts +5 -0
  75. package/dist/cli/CLIOptionsSchema.js +36 -0
  76. package/dist/cli/CLIOptionsSchema.js.map +1 -0
  77. package/dist/cli/CLIParser.d.ts +6 -0
  78. package/dist/cli/CLIParser.js +34 -0
  79. package/dist/cli/CLIParser.js.map +1 -0
  80. package/dist/cli/CLIStarter.d.ts +1 -0
  81. package/dist/cli/CLIStarter.js +3 -0
  82. package/dist/cli/CLIStarter.js.map +1 -0
  83. package/dist/codecs/FFMpegTranscoder.d.ts +20 -0
  84. package/dist/codecs/FFMpegTranscoder.js +169 -0
  85. package/dist/codecs/FFMpegTranscoder.js.map +1 -0
  86. package/dist/codecs/TIMITCodec.d.ts +9 -0
  87. package/dist/codecs/TIMITCodec.js +14 -0
  88. package/dist/codecs/TIMITCodec.js.map +1 -0
  89. package/dist/codecs/WaveCodec.d.ts +19 -0
  90. package/dist/codecs/WaveCodec.js +208 -0
  91. package/dist/codecs/WaveCodec.js.map +1 -0
  92. package/dist/denoising/RNNoise.d.ts +6 -0
  93. package/dist/denoising/RNNoise.js +68 -0
  94. package/dist/denoising/RNNoise.js.map +1 -0
  95. package/dist/dsp/BiquadFilter.d.ts +26 -0
  96. package/dist/dsp/BiquadFilter.js +399 -0
  97. package/dist/dsp/BiquadFilter.js.map +1 -0
  98. package/dist/dsp/FFT.d.ts +9 -0
  99. package/dist/dsp/FFT.js +135 -0
  100. package/dist/dsp/FFT.js.map +1 -0
  101. package/dist/dsp/MFCC.d.ts +25 -0
  102. package/dist/dsp/MFCC.js +162 -0
  103. package/dist/dsp/MFCC.js.map +1 -0
  104. package/dist/dsp/MelSpectogram.d.ts +19 -0
  105. package/dist/dsp/MelSpectogram.js +102 -0
  106. package/dist/dsp/MelSpectogram.js.map +1 -0
  107. package/dist/dsp/Rubberband.d.ts +52 -0
  108. package/dist/dsp/Rubberband.js +186 -0
  109. package/dist/dsp/Rubberband.js.map +1 -0
  110. package/dist/dsp/Sonic.d.ts +2 -0
  111. package/dist/dsp/Sonic.js +39 -0
  112. package/dist/dsp/Sonic.js.map +1 -0
  113. package/dist/dsp/SpeexResampler.d.ts +3 -0
  114. package/dist/dsp/SpeexResampler.js +53 -0
  115. package/dist/dsp/SpeexResampler.js.map +1 -0
  116. package/dist/math/VectorMath.d.ts +70 -0
  117. package/dist/math/VectorMath.js +564 -0
  118. package/dist/math/VectorMath.js.map +1 -0
  119. package/dist/nlp/ChineseSegmentation.d.ts +1 -0
  120. package/dist/nlp/ChineseSegmentation.js +53 -0
  121. package/dist/nlp/ChineseSegmentation.js.map +1 -0
  122. package/dist/nlp/CompromiseNLP.d.ts +15 -0
  123. package/dist/nlp/CompromiseNLP.js +66 -0
  124. package/dist/nlp/CompromiseNLP.js.map +1 -0
  125. package/dist/nlp/EspeakPhonemizer.d.ts +4 -0
  126. package/dist/nlp/EspeakPhonemizer.js +133 -0
  127. package/dist/nlp/EspeakPhonemizer.js.map +1 -0
  128. package/dist/nlp/IPA.d.ts +19 -0
  129. package/dist/nlp/IPA.js +113 -0
  130. package/dist/nlp/IPA.js.map +1 -0
  131. package/dist/nlp/JapaneseSegmentation.d.ts +1 -0
  132. package/dist/nlp/JapaneseSegmentation.js +40 -0
  133. package/dist/nlp/JapaneseSegmentation.js.map +1 -0
  134. package/dist/nlp/Lexicon.d.ts +17 -0
  135. package/dist/nlp/Lexicon.js +6 -0
  136. package/dist/nlp/Lexicon.js.map +1 -0
  137. package/dist/nlp/PhoneConversion.d.ts +6 -0
  138. package/dist/nlp/PhoneConversion.js +467 -0
  139. package/dist/nlp/PhoneConversion.js.map +1 -0
  140. package/dist/nlp/Segmentation.d.ts +41 -0
  141. package/dist/nlp/Segmentation.js +158 -0
  142. package/dist/nlp/Segmentation.js.map +1 -0
  143. package/dist/nlp/TextNormalizer.d.ts +4 -0
  144. package/dist/nlp/TextNormalizer.js +69 -0
  145. package/dist/nlp/TextNormalizer.js.map +1 -0
  146. package/dist/recognition/AmazonTranscribeSTT.d.ts +6 -0
  147. package/dist/recognition/AmazonTranscribeSTT.js +79 -0
  148. package/dist/recognition/AmazonTranscribeSTT.js.map +1 -0
  149. package/dist/recognition/AzureCognitiveServicesSTT.d.ts +7 -0
  150. package/dist/recognition/AzureCognitiveServicesSTT.js +51 -0
  151. package/dist/recognition/AzureCognitiveServicesSTT.js.map +1 -0
  152. package/dist/recognition/GoogleCloudSTT.d.ts +7 -0
  153. package/dist/recognition/GoogleCloudSTT.js +66 -0
  154. package/dist/recognition/GoogleCloudSTT.js.map +1 -0
  155. package/dist/recognition/SileroSTT.d.ts +9 -0
  156. package/dist/recognition/SileroSTT.js +125 -0
  157. package/dist/recognition/SileroSTT.js.map +1 -0
  158. package/dist/recognition/VoskSTT.d.ts +10 -0
  159. package/dist/recognition/VoskSTT.js +78 -0
  160. package/dist/recognition/VoskSTT.js.map +1 -0
  161. package/dist/recognition/WhisperSTT.d.ts +69 -0
  162. package/dist/recognition/WhisperSTT.js +977 -0
  163. package/dist/recognition/WhisperSTT.js.map +1 -0
  164. package/dist/server/Server.d.ts +1 -0
  165. package/dist/server/Server.js +18 -0
  166. package/dist/server/Server.js.map +1 -0
  167. package/dist/speech-language-detection/SileroLanguageDetection.d.ts +9 -0
  168. package/dist/speech-language-detection/SileroLanguageDetection.js +47 -0
  169. package/dist/speech-language-detection/SileroLanguageDetection.js.map +1 -0
  170. package/dist/subtitles/Subtitles.d.ts +25 -0
  171. package/dist/subtitles/Subtitles.js +286 -0
  172. package/dist/subtitles/Subtitles.js.map +1 -0
  173. package/dist/synthesis/AwsPollyTTS.d.ts +7 -0
  174. package/dist/synthesis/AwsPollyTTS.js +51 -0
  175. package/dist/synthesis/AwsPollyTTS.js.map +1 -0
  176. package/dist/synthesis/AzureCognitiveServicesTTS.d.ts +9 -0
  177. package/dist/synthesis/AzureCognitiveServicesTTS.js +103 -0
  178. package/dist/synthesis/AzureCognitiveServicesTTS.js.map +1 -0
  179. package/dist/synthesis/CoquiServerTTS.d.ts +6 -0
  180. package/dist/synthesis/CoquiServerTTS.js +22 -0
  181. package/dist/synthesis/CoquiServerTTS.js.map +1 -0
  182. package/dist/synthesis/ElevenLabsTTS.d.ts +8 -0
  183. package/dist/synthesis/ElevenLabsTTS.js +48 -0
  184. package/dist/synthesis/ElevenLabsTTS.js.map +1 -0
  185. package/dist/synthesis/EspeakTTS.d.ts +46 -0
  186. package/dist/synthesis/EspeakTTS.js +353 -0
  187. package/dist/synthesis/EspeakTTS.js.map +1 -0
  188. package/dist/synthesis/FliteTTS.d.ts +17 -0
  189. package/dist/synthesis/FliteTTS.js +326 -0
  190. package/dist/synthesis/FliteTTS.js.map +1 -0
  191. package/dist/synthesis/GoogleCloudTTS.d.ts +16 -0
  192. package/dist/synthesis/GoogleCloudTTS.js +72 -0
  193. package/dist/synthesis/GoogleCloudTTS.js.map +1 -0
  194. package/dist/synthesis/GoogleTranslateTTS.d.ts +13 -0
  195. package/dist/synthesis/GoogleTranslateTTS.js +177 -0
  196. package/dist/synthesis/GoogleTranslateTTS.js.map +1 -0
  197. package/dist/synthesis/MicrosoftEdgeTTS.d.ts +10 -0
  198. package/dist/synthesis/MicrosoftEdgeTTS.js +216 -0
  199. package/dist/synthesis/MicrosoftEdgeTTS.js.map +1 -0
  200. package/dist/synthesis/SamTTS.d.ts +4 -0
  201. package/dist/synthesis/SamTTS.js +21 -0
  202. package/dist/synthesis/SamTTS.js.map +1 -0
  203. package/dist/synthesis/SapiTTS.d.ts +9 -0
  204. package/dist/synthesis/SapiTTS.js +211 -0
  205. package/dist/synthesis/SapiTTS.js.map +1 -0
  206. package/dist/synthesis/StreamlabsPollyTTS.d.ts +12 -0
  207. package/dist/synthesis/StreamlabsPollyTTS.js +88 -0
  208. package/dist/synthesis/StreamlabsPollyTTS.js.map +1 -0
  209. package/dist/synthesis/SvoxPicoTTS.d.ts +11 -0
  210. package/dist/synthesis/SvoxPicoTTS.js +236 -0
  211. package/dist/synthesis/SvoxPicoTTS.js.map +1 -0
  212. package/dist/synthesis/VitsTTS.d.ts +27 -0
  213. package/dist/synthesis/VitsTTS.js +359 -0
  214. package/dist/synthesis/VitsTTS.js.map +1 -0
  215. package/dist/tests/Test.d.ts +1 -0
  216. package/dist/tests/Test.js +10 -0
  217. package/dist/tests/Test.js.map +1 -0
  218. package/dist/text-language-detection/FastTextLanguageDetection.d.ts +2 -0
  219. package/dist/text-language-detection/FastTextLanguageDetection.js +38 -0
  220. package/dist/text-language-detection/FastTextLanguageDetection.js.map +1 -0
  221. package/dist/text-language-detection/TinyLDLanguageDetection.d.ts +2 -0
  222. package/dist/text-language-detection/TinyLDLanguageDetection.js +12 -0
  223. package/dist/text-language-detection/TinyLDLanguageDetection.js.map +1 -0
  224. package/dist/utilities/BinaryArrayConversion.d.ts +15 -0
  225. package/dist/utilities/BinaryArrayConversion.js +115 -0
  226. package/dist/utilities/BinaryArrayConversion.js.map +1 -0
  227. package/dist/utilities/Compression.d.ts +4 -0
  228. package/dist/utilities/Compression.js +67 -0
  229. package/dist/utilities/Compression.js.map +1 -0
  230. package/dist/utilities/FileDownloader.d.ts +3 -0
  231. package/dist/utilities/FileDownloader.js +147 -0
  232. package/dist/utilities/FileDownloader.js.map +1 -0
  233. package/dist/utilities/FileSystem.d.ts +30 -0
  234. package/dist/utilities/FileSystem.js +141 -0
  235. package/dist/utilities/FileSystem.js.map +1 -0
  236. package/dist/utilities/Hashing.d.ts +10 -0
  237. package/dist/utilities/Hashing.js +169 -0
  238. package/dist/utilities/Hashing.js.map +1 -0
  239. package/dist/utilities/Locale.d.ts +9 -0
  240. package/dist/utilities/Locale.js +65 -0
  241. package/dist/utilities/Locale.js.map +1 -0
  242. package/dist/utilities/Logger.d.ts +11 -0
  243. package/dist/utilities/Logger.js +49 -0
  244. package/dist/utilities/Logger.js.map +1 -0
  245. package/dist/utilities/NdArrayUtilities.d.ts +3 -0
  246. package/dist/utilities/NdArrayUtilities.js +22 -0
  247. package/dist/utilities/NdArrayUtilities.js.map +1 -0
  248. package/dist/utilities/ObjectUtilities.d.ts +4 -0
  249. package/dist/utilities/ObjectUtilities.js +132 -0
  250. package/dist/utilities/ObjectUtilities.js.map +1 -0
  251. package/dist/utilities/OpenPromise.d.ts +6 -0
  252. package/dist/utilities/OpenPromise.js +12 -0
  253. package/dist/utilities/OpenPromise.js.map +1 -0
  254. package/dist/utilities/PackageManager.d.ts +4 -0
  255. package/dist/utilities/PackageManager.js +46 -0
  256. package/dist/utilities/PackageManager.js.map +1 -0
  257. package/dist/utilities/RandomGenerator.d.ts +35 -0
  258. package/dist/utilities/RandomGenerator.js +149 -0
  259. package/dist/utilities/RandomGenerator.js.map +1 -0
  260. package/dist/utilities/TarballMaker.d.ts +4 -0
  261. package/dist/utilities/TarballMaker.js +50 -0
  262. package/dist/utilities/TarballMaker.js.map +1 -0
  263. package/dist/utilities/Timeline.d.ts +20 -0
  264. package/dist/utilities/Timeline.js +110 -0
  265. package/dist/utilities/Timeline.js.map +1 -0
  266. package/dist/utilities/Timer.d.ts +13 -0
  267. package/dist/utilities/Timer.js +70 -0
  268. package/dist/utilities/Timer.js.map +1 -0
  269. package/dist/utilities/Utilities.d.ts +68 -0
  270. package/dist/utilities/Utilities.js +305 -0
  271. package/dist/utilities/Utilities.js.map +1 -0
  272. package/dist/utilities/WasmMemoryManager.d.ts +142 -0
  273. package/dist/utilities/WasmMemoryManager.js +407 -0
  274. package/dist/utilities/WasmMemoryManager.js.map +1 -0
  275. package/dist/utilities/WebReader.d.ts +1 -0
  276. package/dist/utilities/WebReader.js +47 -0
  277. package/dist/utilities/WebReader.js.map +1 -0
  278. package/dist/utilities/WikipediaReader.d.ts +1 -0
  279. package/dist/utilities/WikipediaReader.js +31 -0
  280. package/dist/utilities/WikipediaReader.js.map +1 -0
  281. package/dist/voice-activity-detection/SileroVAD.d.ts +13 -0
  282. package/dist/voice-activity-detection/SileroVAD.js +58 -0
  283. package/dist/voice-activity-detection/SileroVAD.js.map +1 -0
  284. package/dist/voice-activity-detection/WebRtcVAD.d.ts +3 -0
  285. package/dist/voice-activity-detection/WebRtcVAD.js +53 -0
  286. package/dist/voice-activity-detection/WebRtcVAD.js.map +1 -0
  287. package/docs/CLI.md +234 -0
  288. package/docs/Development.md +3 -0
  289. package/docs/Engines.md +80 -0
  290. package/docs/Licenses.md +37 -0
  291. package/docs/Options.md +192 -0
  292. package/docs/Roadmap.md +16 -0
  293. package/docs/Technical.md +72 -0
  294. package/package.json +119 -19
  295. package/cli.js +0 -3
@@ -0,0 +1,216 @@
1
+ import { request } from "gaxios";
2
+ import WebSocket from "ws";
3
+ import { escape } from 'html-escaper';
4
+ import { randomUUID } from "crypto";
5
+ import splitBuffer from "buffer-split";
6
+ import * as AzureCognitiveServicesTTS from "./AzureCognitiveServicesTTS.js";
7
+ import * as FFMpegTranscoder from "../codecs/FFMpegTranscoder.js";
8
+ import { Logger } from "../utilities/Logger.js";
9
+ import { OpenPromise } from "../utilities/OpenPromise.js";
10
+ import { logToStderr } from "../utilities/Utilities.js";
11
+ import { getRawAudioDuration } from "../audio/AudioUtilities.js";
12
+ const traceEnabled = false;
13
+ const log = traceEnabled ? logToStderr : () => { };
14
+ export async function synthesize(text, trustedClientToken, voice = "Microsoft Server Speech Text to Speech Voice (en-US, ChristopherNeural)", ssmlPitchString = "+0Hz", ssmlRateString = "+0%", ssmlVolumeString = "+0%") {
15
+ const logger = new Logger();
16
+ logger.start("Request synthesis from Microsoft Edge cloud API");
17
+ const { audioData, events } = await requestSynthesis(text, trustedClientToken, voice, ssmlPitchString, ssmlRateString, ssmlVolumeString);
18
+ logger.end();
19
+ const rawAudio = await FFMpegTranscoder.decodeToChannels(audioData, 24000, 1);
20
+ logger.start("Convert boundary events to timeline");
21
+ const timeline = AzureCognitiveServicesTTS.boundaryEventsToTimeline(events, getRawAudioDuration(rawAudio));
22
+ logger.end();
23
+ return { rawAudio, timeline };
24
+ }
25
+ async function requestSynthesis(text, trustedClientToken, voice = "Microsoft Server Speech Text to Speech Voice (en-US, ChristopherNeural)", ssmlPitchString = "+0Hz", ssmlRateString = "+0%", ssmlVolumeString = "+0%") {
26
+ const synthesisOpenPromise = new OpenPromise();
27
+ const requestId = getRandomID();
28
+ const webSocket = await initializeWebsocketConnection(trustedClientToken);
29
+ const receivedBinaryMessages = [];
30
+ const receivedEventMessages = [];
31
+ const audioChunks = [];
32
+ const onMessage = (messageData, isBinary) => {
33
+ if (isBinary) {
34
+ const [header, audioChunk] = parseBinaryMessage(messageData);
35
+ if (header['X-RequestId'] != requestId) {
36
+ log(`Ignoring binary message to different request Id: ${requestId}`);
37
+ return;
38
+ }
39
+ log("\nReceived binary message:");
40
+ log(header);
41
+ receivedBinaryMessages.push(messageData);
42
+ receivedBinaryMessages.push(Buffer.from("\n\n"));
43
+ audioChunks.push(audioChunk);
44
+ }
45
+ else {
46
+ const message = messageData.toString("utf8");
47
+ const [header, content] = parseTextMessage(message);
48
+ if (header['X-RequestId'] != requestId) {
49
+ log(`Ignoring text message to different request Id: ${requestId}`);
50
+ return;
51
+ }
52
+ log("\nReceived text message:");
53
+ log(header);
54
+ log(content);
55
+ if (header["Path"] == "turn.start") {
56
+ }
57
+ else if (header["Path"] == "audio.metadata") {
58
+ receivedEventMessages.push(content["Metadata"][0]);
59
+ }
60
+ else if (header["Path"] == "turn.end") {
61
+ const result = { audioData: Buffer.concat(audioChunks), events: receivedEventMessages };
62
+ webSocket.off("message", onMessage);
63
+ webSocket.off("error", onError);
64
+ webSocket.off("close", onClose);
65
+ synthesisOpenPromise.resolve(result);
66
+ }
67
+ }
68
+ };
69
+ const onError = (err) => {
70
+ synthesisOpenPromise.reject(err);
71
+ };
72
+ const onClose = async (code, reason) => {
73
+ log("WebSocket closed.");
74
+ log(code);
75
+ log(reason.toString());
76
+ if (reason.length > 0) {
77
+ synthesisOpenPromise.reject(`Websocket closed with code ${code}, reason: ${reason}`);
78
+ }
79
+ };
80
+ webSocket.on("message", onMessage);
81
+ webSocket.on("error", onError);
82
+ webSocket.on("close", onClose);
83
+ const requestContentSSML = `<speak version='1.0' xmlns='http://www.w3.org/2001/10/synthesis' xml:lang='en-US'>` +
84
+ `<voice name='${voice}'>` +
85
+ `<prosody pitch='${ssmlPitchString}' rate='${ssmlRateString}' volume='${ssmlVolumeString}'>` +
86
+ escape(text) +
87
+ `</prosody>` +
88
+ `</voice>` +
89
+ `</speak>`;
90
+ const bodyRequestString = `X-RequestId:${requestId}\r\n` +
91
+ `Content-Type:application/ssml+xml\r\n` +
92
+ `X-Timestamp:${getTimestampString()}Z\r\n` + // The added "Z" recreates a Microsoft Edge bug.
93
+ `Path:ssml\r\n\r\n` +
94
+ requestContentSSML;
95
+ log(bodyRequestString);
96
+ webSocket.send(bodyRequestString);
97
+ return synthesisOpenPromise.promise;
98
+ }
99
+ let existingWebSocketConnection = undefined;
100
+ export async function initializeWebsocketConnection(trustedClientToken) {
101
+ const requestOpenPromise = new OpenPromise();
102
+ if (existingWebSocketConnection && existingWebSocketConnection.readyState == WebSocket.OPEN) {
103
+ log(`Existing websocket connection is still open. Reusing it.`);
104
+ requestOpenPromise.resolve(existingWebSocketConnection);
105
+ return requestOpenPromise.promise;
106
+ }
107
+ else {
108
+ existingWebSocketConnection = undefined;
109
+ }
110
+ const connectionId = getRandomID();
111
+ const requestURL = "wss://speech.platform.bing.com/" +
112
+ "consumer/speech/synthesize/readaloud/edge/v1" +
113
+ `?TrustedClientToken=${trustedClientToken}` +
114
+ "&ConnectionId=" + connectionId;
115
+ log(requestURL);
116
+ log("");
117
+ const webSocket = new WebSocket(requestURL, {
118
+ headers: {
119
+ "Pragma": "no-cache",
120
+ "Cache-Control": "no-cache",
121
+ "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/103.0.5060.66 Safari/537.36 Edg/103.0.1264.44",
122
+ "Origin": "chrome-extension://jdiccldimpdaibmpdkjnbmckianbfold",
123
+ "Accept-Encoding": "gzip, deflate, br",
124
+ "Accept-Language": "en-US,en;q=0.9",
125
+ }
126
+ });
127
+ const requestMetadata = {
128
+ "context": {
129
+ "synthesis": {
130
+ "audio": {
131
+ "metadataoptions": {
132
+ "sentenceBoundaryEnabled": "false",
133
+ "wordBoundaryEnabled": "true"
134
+ },
135
+ "outputFormat": "webm-24khz-16bit-mono-opus"
136
+ }
137
+ }
138
+ }
139
+ };
140
+ const onOpen = () => {
141
+ const configRequestString = `X-Timestamp:${getTimestampString()}\r\n` +
142
+ `Content-Type:application/json; charset=utf-8\r\n` +
143
+ `Path:speech.config\r\n\r\n` +
144
+ `${JSON.stringify(requestMetadata, null, 0)}\r\n`;
145
+ log(configRequestString);
146
+ webSocket.send(configRequestString);
147
+ existingWebSocketConnection = webSocket;
148
+ webSocket.off("open", onOpen);
149
+ webSocket.off("close", onClose);
150
+ webSocket.off("error", onError);
151
+ requestOpenPromise.resolve(webSocket);
152
+ };
153
+ const onClose = (code, reason) => {
154
+ log("WebSocket closed.");
155
+ log(code);
156
+ log(reason.toString());
157
+ if (reason.length > 0) {
158
+ requestOpenPromise.reject(`Websocket closed with code ${code}, reason: ${reason}`);
159
+ }
160
+ };
161
+ const onError = (err) => {
162
+ requestOpenPromise.reject(err);
163
+ };
164
+ webSocket.on("open", onOpen);
165
+ webSocket.on("close", onClose);
166
+ webSocket.on("error", onError);
167
+ return requestOpenPromise.promise;
168
+ }
169
+ export async function getVoiceList(trustedClientToken) {
170
+ const response = await request({
171
+ method: "GET",
172
+ url: "https://speech.platform.bing.com/consumer/speech/synthesize/" +
173
+ `readaloud/voices/list?trustedclienttoken=${trustedClientToken}`,
174
+ headers: {
175
+ "sec-ch-ua": `" Not;A Brand";v="99", "Microsoft Edge";v="103", "Chromium";v="103"`,
176
+ "dnt": "1",
177
+ "sec-ch-ua-mobile": "?0",
178
+ "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/103.0.5060.66 Safari/537.36 Edg/103.0.1264.44",
179
+ "sec-ch-ua-platform": "Windows",
180
+ "accept": "*/*",
181
+ "sec-fetch-site": "none",
182
+ "sec-fetch-mode": "cors",
183
+ "sec-fetch-dest": "empty",
184
+ "accept-encoding": "gzip, deflate, br",
185
+ "accept-language": "en-US,en;q=0.9",
186
+ },
187
+ responseType: "json"
188
+ });
189
+ return response.data;
190
+ }
191
+ function getTimestampString() {
192
+ const timestampString = new Date(new Date().toLocaleString("en-US", { timeZone: "UTC" })).toString().replace(/GMT.*/, "GMT+0000 (Coordinated Universal Time)");
193
+ return timestampString;
194
+ }
195
+ function getRandomID() {
196
+ return randomUUID().replace(/-/g, "");
197
+ }
198
+ function parseHeaderString(headerString) {
199
+ const headers = {};
200
+ for (const headerLine of headerString.split("\r\n")) {
201
+ const [key, value] = headerLine.split(":");
202
+ headers[key] = value;
203
+ }
204
+ return headers;
205
+ }
206
+ function parseTextMessage(message) {
207
+ const [headerString, content] = message.split("\r\n\r\n");
208
+ return [parseHeaderString(headerString), JSON.parse(content)];
209
+ }
210
+ function parseBinaryMessage(message) {
211
+ const audioChunk = splitBuffer(message, Buffer.from("Path:audio\r\n"))[1];
212
+ const headerBuffer = message.subarray(2, message.length - audioChunk.length);
213
+ const headerString = headerBuffer.toString("utf8").trim();
214
+ return [parseHeaderString(headerString), audioChunk];
215
+ }
216
+ //# sourceMappingURL=MicrosoftEdgeTTS.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"MicrosoftEdgeTTS.js","sourceRoot":"","sources":["../../src/synthesis/MicrosoftEdgeTTS.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,OAAO,EAAE,MAAM,QAAQ,CAAA;AAChC,OAAO,SAAS,MAAM,IAAI,CAAA;AAE1B,OAAO,EAAE,MAAM,EAAE,MAAM,cAAc,CAAA;AAErC,OAAO,EAAE,UAAU,EAAE,MAAM,QAAQ,CAAA;AACnC,OAAO,WAAW,MAAM,cAAc,CAAA;AAEtC,OAAO,KAAK,yBAAyB,MAAM,gCAAgC,CAAA;AAC3E,OAAO,KAAK,gBAAgB,MAAM,+BAA+B,CAAA;AACjE,OAAO,EAAE,MAAM,EAAE,MAAM,wBAAwB,CAAA;AAC/C,OAAO,EAAE,WAAW,EAAE,MAAM,6BAA6B,CAAA;AACzD,OAAO,EAAE,WAAW,EAAE,MAAM,2BAA2B,CAAA;AACvD,OAAO,EAAE,mBAAmB,EAAE,MAAM,4BAA4B,CAAA;AAEhE,MAAM,YAAY,GAAG,KAAK,CAAA;AAE1B,MAAM,GAAG,GAAuB,YAAY,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAA;AAEtE,MAAM,CAAC,KAAK,UAAU,UAAU,CAC/B,IAAY,EACZ,kBAA0B,EAC1B,KAAK,GAAG,yEAAyE,EACjF,eAAe,GAAG,MAAM,EACxB,cAAc,GAAG,KAAK,EACtB,gBAAgB,GAAG,KAAK;IAExB,MAAM,MAAM,GAAG,IAAI,MAAM,EAAE,CAAA;IAC3B,MAAM,CAAC,KAAK,CAAC,iDAAiD,CAAC,CAAA;IAE/D,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,GAAG,MAAM,gBAAgB,CAAC,IAAI,EAAE,kBAAkB,EAAE,KAAK,EAAE,eAAe,EAAE,cAAc,EAAE,gBAAgB,CAAC,CAAA;IACxI,MAAM,CAAC,GAAG,EAAE,CAAA;IAEZ,MAAM,QAAQ,GAAG,MAAM,gBAAgB,CAAC,gBAAgB,CAAC,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC,CAAA;IAE7E,MAAM,CAAC,KAAK,CAAC,qCAAqC,CAAC,CAAA;IACnD,MAAM,QAAQ,GAAG,yBAAyB,CAAC,wBAAwB,CAAC,MAAM,EAAE,mBAAmB,CAAC,QAAQ,CAAC,CAAC,CAAA;IAC1G,MAAM,CAAC,GAAG,EAAE,CAAA;IAEZ,OAAO,EAAE,QAAQ,EAAE,QAAQ,EAAE,CAAA;AAC9B,CAAC;AAID,KAAK,UAAU,gBAAgB,CAC9B,IAAY,EACZ,kBAA0B,EAC1B,KAAK,GAAG,yEAAyE,EACjF,eAAe,GAAG,MAAM,EACxB,cAAc,GAAG,KAAK,EACtB,gBAAgB,GAAG,KAAK;IAExB,MAAM,oBAAoB,GAAG,IAAI,WAAW,EAA0B,CAAA;IAEtE,MAAM,SAAS,GAAG,WAAW,EAAE,CAAA;IAE/B,MAAM,SAAS,GAAG,MAAM,6BAA6B,CAAC,kBAAkB,CAAC,CAAA;IAEzE,MAAM,sBAAsB,GAAa,EAAE,CAAA;IAC3C,MAAM,qBAAqB,GAAU,EAAE,CAAA;IACvC,MAAM,WAAW,GAAa,EAAE,CAAA;IAEhC,MAAM,SAAS,GAAG,CAAC,WAAmB,EAAE,QAAiB,EAAE,EAAE;QAC5D,IAAI,QAAQ,EAAE;YACb,MAAM,CAAC,MAAM,EAAE,UAAU,CAAC,GAAG,kBAAkB,CAAC,WAAW,CAAC,CAAA;YAE5D,IAAI,MAAM,CAAC,aAAa,CAAC,IAAI,SAAS,EAAE;gBACvC,GAAG,CAAC,oDAAoD,SAAS,EAAE,CAAC,CAAA;gBAEpE,OAAM;aACN;YAED,GAAG,CAAC,4BAA4B,CAAC,CAAA;YACjC,GAAG,CAAC,MAAM,CAAC,CAAA;YAEX,sBAAsB,CAAC,IAAI,CAAC,WAAW,CAAC,CAAA;YACxC,sBAAsB,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAA;YAEhD,WAAW,CAAC,IAAI,CAAC,UAAU,CAAC,CAAA;SAC5B;aAAM;YACN,MAAM,OAAO,GAAG,WAAW,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAA;YAC5C,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,gBAAgB,CAAC,OAAO,CAAC,CAAA;YAEnD,IAAI,MAAM,CAAC,aAAa,CAAC,IAAI,SAAS,EAAE;gBACvC,GAAG,CAAC,kDAAkD,SAAS,EAAE,CAAC,CAAA;gBAClE,OAAM;aACN;YAED,GAAG,CAAC,0BAA0B,CAAC,CAAA;YAC/B,GAAG,CAAC,MAAM,CAAC,CAAA;YACX,GAAG,CAAC,OAAO,CAAC,CAAA;YAEZ,IAAI,MAAM,CAAC,MAAM,CAAC,IAAI,YAAY,EAAE;aAEnC;iBACI,IAAI,MAAM,CAAC,MAAM,CAAC,IAAI,gBAAgB,EAAE;gBAC5C,qBAAqB,CAAC,IAAI,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,CAAC,CAAA;aAClD;iBACI,IAAI,MAAM,CAAC,MAAM,CAAC,IAAI,UAAU,EAAE;gBACtC,MAAM,MAAM,GAA2B,EAAE,SAAS,EAAE,MAAM,CAAC,MAAM,CAAC,WAAW,CAAC,EAAE,MAAM,EAAE,qBAAqB,EAAE,CAAA;gBAE/G,SAAS,CAAC,GAAG,CAAC,SAAS,EAAE,SAAS,CAAC,CAAA;gBACnC,SAAS,CAAC,GAAG,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;gBAC/B,SAAS,CAAC,GAAG,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;gBAE/B,oBAAoB,CAAC,OAAO,CAAC,MAAM,CAAC,CAAA;aACpC;SACD;IACF,CAAC,CAAA;IAED,MAAM,OAAO,GAAG,CAAC,GAAU,EAAE,EAAE;QAC9B,oBAAoB,CAAC,MAAM,CAAC,GAAG,CAAC,CAAA;IACjC,CAAC,CAAA;IAED,MAAM,OAAO,GAAG,KAAK,EAAE,IAAY,EAAE,MAAc,EAAE,EAAE;QACtD,GAAG,CAAC,mBAAmB,CAAC,CAAA;QACxB,GAAG,CAAC,IAAI,CAAC,CAAA;QACT,GAAG,CAAC,MAAM,CAAC,QAAQ,EAAE,CAAC,CAAA;QAEtB,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE;YACtB,oBAAoB,CAAC,MAAM,CAAC,8BAA8B,IAAI,aAAa,MAAM,EAAE,CAAC,CAAA;SACpF;IACF,CAAC,CAAA;IAED,SAAS,CAAC,EAAE,CAAC,SAAS,EAAE,SAAS,CAAC,CAAA;IAClC,SAAS,CAAC,EAAE,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;IAC9B,SAAS,CAAC,EAAE,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;IAE9B,MAAM,kBAAkB,GACvB,oFAAoF;QACpF,iBAAiB,KAAK,IAAI;QAC1B,mBAAmB,eAAe,WAAW,cAAc,aAAa,gBAAgB,IAAI;QAC5F,MAAM,CAAC,IAAI,CAAC;QACZ,YAAY;QACZ,UAAU;QACV,UAAU,CAAA;IAEX,MAAM,iBAAiB,GACtB,eAAe,SAAS,MAAM;QAC9B,uCAAuC;QACvC,eAAe,kBAAkB,EAAE,OAAO,GAAG,gDAAgD;QAC7F,mBAAmB;QACnB,kBAAkB,CAAA;IAEnB,GAAG,CAAC,iBAAiB,CAAC,CAAA;IAEtB,SAAS,CAAC,IAAI,CAAC,iBAAiB,CAAC,CAAA;IAEjC,OAAO,oBAAoB,CAAC,OAAO,CAAA;AACpC,CAAC;AAED,IAAI,2BAA2B,GAA0B,SAAS,CAAA;AAElE,MAAM,CAAC,KAAK,UAAU,6BAA6B,CAAC,kBAA0B;IAC7E,MAAM,kBAAkB,GAAG,IAAI,WAAW,EAAa,CAAA;IAEvD,IAAI,2BAA2B,IAAI,2BAA2B,CAAC,UAAU,IAAI,SAAS,CAAC,IAAI,EAAE;QAC5F,GAAG,CAAC,0DAA0D,CAAC,CAAA;QAE/D,kBAAkB,CAAC,OAAO,CAAC,2BAA2B,CAAC,CAAA;QAEvD,OAAO,kBAAkB,CAAC,OAAO,CAAA;KACjC;SAAM;QACN,2BAA2B,GAAG,SAAS,CAAA;KACvC;IAED,MAAM,YAAY,GAAG,WAAW,EAAE,CAAA;IAElC,MAAM,UAAU,GAAG,iCAAiC;QACnD,8CAA8C;QAC9C,uBAAuB,kBAAkB,EAAE;QAC3C,gBAAgB,GAAG,YAAY,CAAA;IAEhC,GAAG,CAAC,UAAU,CAAC,CAAA;IACf,GAAG,CAAC,EAAE,CAAC,CAAA;IAEP,MAAM,SAAS,GAAG,IAAI,SAAS,CAAC,UAAU,EAAE;QAC3C,OAAO,EAAE;YACR,QAAQ,EAAE,UAAU;YACpB,eAAe,EAAE,UAAU;YAC3B,YAAY,EAAE,uIAAuI;YACrJ,QAAQ,EAAE,qDAAqD;YAC/D,iBAAiB,EAAE,mBAAmB;YACtC,iBAAiB,EAAE,gBAAgB;SACnC;KACD,CAAC,CAAA;IAEF,MAAM,eAAe,GAAG;QACvB,SAAS,EAAE;YACV,WAAW,EAAE;gBACZ,OAAO,EAAE;oBACR,iBAAiB,EAAE;wBAClB,yBAAyB,EAAE,OAAO;wBAClC,qBAAqB,EAAE,MAAM;qBAC7B;oBACD,cAAc,EAAE,4BAA4B;iBAC5C;aACD;SACD;KACD,CAAA;IAED,MAAM,MAAM,GAAG,GAAG,EAAE;QACnB,MAAM,mBAAmB,GACxB,eAAe,kBAAkB,EAAE,MAAM;YACzC,kDAAkD;YAClD,4BAA4B;YAC5B,GAAG,IAAI,CAAC,SAAS,CAAC,eAAe,EAAE,IAAI,EAAE,CAAC,CAAC,MAAM,CAAA;QAElD,GAAG,CAAC,mBAAmB,CAAC,CAAA;QAExB,SAAS,CAAC,IAAI,CAAC,mBAAmB,CAAC,CAAA;QAEnC,2BAA2B,GAAG,SAAS,CAAA;QAEvC,SAAS,CAAC,GAAG,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;QAC7B,SAAS,CAAC,GAAG,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;QAC/B,SAAS,CAAC,GAAG,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;QAE/B,kBAAkB,CAAC,OAAO,CAAC,SAAS,CAAC,CAAA;IACtC,CAAC,CAAA;IAED,MAAM,OAAO,GAAG,CAAC,IAAY,EAAE,MAAc,EAAE,EAAE;QAChD,GAAG,CAAC,mBAAmB,CAAC,CAAA;QACxB,GAAG,CAAC,IAAI,CAAC,CAAA;QACT,GAAG,CAAC,MAAM,CAAC,QAAQ,EAAE,CAAC,CAAA;QAEtB,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE;YACtB,kBAAkB,CAAC,MAAM,CAAC,8BAA8B,IAAI,aAAa,MAAM,EAAE,CAAC,CAAA;SAClF;IACF,CAAC,CAAA;IAED,MAAM,OAAO,GAAG,CAAC,GAAU,EAAE,EAAE;QAC9B,kBAAkB,CAAC,MAAM,CAAC,GAAG,CAAC,CAAA;IAC/B,CAAC,CAAA;IAED,SAAS,CAAC,EAAE,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;IAC5B,SAAS,CAAC,EAAE,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;IAC9B,SAAS,CAAC,EAAE,CAAC,OAAO,EAAE,OAAO,CAAC,CAAA;IAE9B,OAAO,kBAAkB,CAAC,OAAO,CAAA;AAClC,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,YAAY,CAAC,kBAA0B;IAC5D,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAM;QACnC,MAAM,EAAE,KAAK;QAEb,GAAG,EAAE,8DAA8D;YAClE,4CAA4C,kBAAkB,EAAE;QAEjE,OAAO,EAAE;YACR,WAAW,EAAE,qEAAqE;YAClF,KAAK,EAAE,GAAG;YACV,kBAAkB,EAAE,IAAI;YACxB,YAAY,EAAE,uIAAuI;YACrJ,oBAAoB,EAAE,SAAS;YAC/B,QAAQ,EAAE,KAAK;YACf,gBAAgB,EAAE,MAAM;YACxB,gBAAgB,EAAE,MAAM;YACxB,gBAAgB,EAAE,OAAO;YACzB,iBAAiB,EAAE,mBAAmB;YACtC,iBAAiB,EAAE,gBAAgB;SACnC;QAED,YAAY,EAAE,MAAM;KACpB,CAAC,CAAA;IAEF,OAAO,QAAQ,CAAC,IAAa,CAAA;AAC9B,CAAC;AACD,SAAS,kBAAkB;IAC1B,MAAM,eAAe,GAAG,IAAI,IAAI,CAAC,IAAI,IAAI,EAAE,CAAC,cAAc,CAAC,OAAO,EAAE,EAAE,QAAQ,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,QAAQ,EAAE,CAAC,OAAO,CAAC,OAAO,EAAE,uCAAuC,CAAC,CAAA;IAC9J,OAAO,eAAe,CAAA;AACvB,CAAC;AAED,SAAS,WAAW;IACnB,OAAO,UAAU,EAAE,CAAC,OAAO,CAAC,IAAI,EAAE,EAAE,CAAC,CAAA;AACtC,CAAC;AAED,SAAS,iBAAiB,CAAC,YAAoB;IAC9C,MAAM,OAAO,GAAQ,EAAE,CAAA;IAEvB,KAAK,MAAM,UAAU,IAAI,YAAY,CAAC,KAAK,CAAC,MAAM,CAAC,EAAE;QACpD,MAAM,CAAC,GAAG,EAAE,KAAK,CAAC,GAAG,UAAU,CAAC,KAAK,CAAC,GAAG,CAAC,CAAA;QAC1C,OAAO,CAAC,GAAG,CAAC,GAAG,KAAK,CAAA;KACpB;IAED,OAAO,OAAO,CAAA;AACf,CAAC;AAED,SAAS,gBAAgB,CAAC,OAAe;IACxC,MAAM,CAAC,YAAY,EAAE,OAAO,CAAC,GAAG,OAAO,CAAC,KAAK,CAAC,UAAU,CAAC,CAAA;IAEzD,OAAO,CAAC,iBAAiB,CAAC,YAAY,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAA;AAC9D,CAAC;AAED,SAAS,kBAAkB,CAAC,OAAe;IAC1C,MAAM,UAAU,GAAG,WAAW,CAAC,OAAO,EAAE,MAAM,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAA;IACzE,MAAM,YAAY,GAAG,OAAO,CAAC,QAAQ,CAAC,CAAC,EAAE,OAAO,CAAC,MAAM,GAAG,UAAU,CAAC,MAAM,CAAC,CAAA;IAC5E,MAAM,YAAY,GAAG,YAAY,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,IAAI,EAAE,CAAA;IAEzD,OAAO,CAAC,iBAAiB,CAAC,YAAY,CAAC,EAAE,UAAU,CAAC,CAAA;AACrD,CAAC"}
@@ -0,0 +1,4 @@
1
+ import { RawAudio } from '../audio/AudioUtilities.js';
2
+ export declare function synthesize(text: string, pitch?: number, speed?: number, mouth?: number, throat?: number): Promise<{
3
+ rawAudio: RawAudio;
4
+ }>;
@@ -0,0 +1,21 @@
1
+ import { Logger } from '../utilities/Logger.js';
2
+ // SAM (Software Automatic Mouth) speech synthesizer from 1982 (pure JS port)
3
+ //
4
+ // https://github.com/discordier/sam
5
+ //
6
+ // https://habr-com.translate.goog/ru/post/500764/?_x_tr_sl=auto&_x_tr_tl=en
7
+ export async function synthesize(text, pitch = 64, speed = 72, mouth = 128, throat = 128) {
8
+ const logger = new Logger();
9
+ logger.start("Initialize sam module");
10
+ const { default: SamJs } = await import('sam-js');
11
+ const sam = new SamJs({ pitch, speed, mouth, throat });
12
+ logger.start("Synthesize with sam");
13
+ const samples = sam.buf32(text);
14
+ if (!samples) {
15
+ throw new Error("Sam TTS failed");
16
+ }
17
+ const rawAudio = { audioChannels: [samples], sampleRate: 22050 };
18
+ logger.end();
19
+ return { rawAudio };
20
+ }
21
+ //# sourceMappingURL=SamTTS.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"SamTTS.js","sourceRoot":"","sources":["../../src/synthesis/SamTTS.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,MAAM,EAAE,MAAM,wBAAwB,CAAA;AAE/C,6EAA6E;AAC7E,EAAE;AACF,oCAAoC;AACpC,EAAE;AACF,4EAA4E;AAE5E,MAAM,CAAC,KAAK,UAAU,UAAU,CAAC,IAAY,EAAE,KAAK,GAAG,EAAE,EAAE,KAAK,GAAG,EAAE,EAAE,KAAK,GAAG,GAAG,EAAE,MAAM,GAAG,GAAG;IAC/F,MAAM,MAAM,GAAG,IAAI,MAAM,EAAE,CAAA;IAC3B,MAAM,CAAC,KAAK,CAAC,uBAAuB,CAAC,CAAA;IAErC,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,QAAQ,CAAC,CAAA;IAEjD,MAAM,GAAG,GAAG,IAAI,KAAK,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,CAAC,CAAA;IAEtD,MAAM,CAAC,KAAK,CAAC,qBAAqB,CAAC,CAAA;IACnC,MAAM,OAAO,GAAiB,GAAG,CAAC,KAAK,CAAC,IAAI,CAAiB,CAAA;IAE7D,IAAI,CAAC,OAAO,EAAE;QACb,MAAM,IAAI,KAAK,CAAC,gBAAgB,CAAC,CAAA;KACjC;IAED,MAAM,QAAQ,GAAa,EAAE,aAAa,EAAE,CAAC,OAAO,CAAC,EAAE,UAAU,EAAE,KAAK,EAAE,CAAA;IAE1E,MAAM,CAAC,GAAG,EAAE,CAAA;IAEZ,OAAO,EAAE,QAAQ,EAAE,CAAA;AACpB,CAAC"}
@@ -0,0 +1,9 @@
1
+ import { SynthesisVoice } from "../api/API.js";
2
+ import { RawAudio } from "../audio/AudioUtilities.js";
3
+ import { Timeline } from "../utilities/Timeline.js";
4
+ export declare function synthesize(text: string, voiceName: string, rate?: number, useSpeechPlatform?: boolean): Promise<{
5
+ rawAudio: RawAudio;
6
+ timeline: Timeline;
7
+ }>;
8
+ export declare function getVoiceList(useSpeechPlatform?: boolean): Promise<SynthesisVoice[]>;
9
+ export declare function AssertSAPIAvailable(testForSpeechPlatform?: boolean): Promise<void>;
@@ -0,0 +1,211 @@
1
+ import { decodeToChannels } from "../audio/AudioBufferConversion.js";
2
+ import { SampleFormat } from "../codecs/WaveCodec.js";
3
+ import { getShortLanguageCode, lcidToIsoLanguageCode } from "../utilities/Locale.js";
4
+ import { Logger } from "../utilities/Logger.js";
5
+ import { logToStderr } from "../utilities/Utilities.js";
6
+ const log = logToStderr;
7
+ export function synthesize(text, voiceName, rate = 0, useSpeechPlatform = false) {
8
+ return new Promise(async (resolve, reject) => {
9
+ const logger = new Logger();
10
+ logger.start("Initialize winax module");
11
+ const { default: WinAX } = await import("winax");
12
+ logger.start("Create SAPI COM object");
13
+ const sapiVoice = new global.ActiveXObject(useSpeechPlatform ? "Speech.SPVoice" : "SAPI.SPVoice");
14
+ sapiVoice.EventInterests = 33790;
15
+ logger.start("Get SAPI voice list and select best match");
16
+ if (voiceName) {
17
+ const voiceObjects = sapiVoice.GetVoices();
18
+ for (let i = 0; i < voiceObjects.Count; i++) {
19
+ const voiceObject = voiceObjects.Item(i);
20
+ const candidateVoiceName = voiceObject.GetDescription();
21
+ if (candidateVoiceName == voiceName) {
22
+ sapiVoice.Voice = voiceObject;
23
+ }
24
+ }
25
+ }
26
+ sapiVoice.Rate = rate;
27
+ // Create phone converter for language
28
+ const sapiPhoneConverter = new global.ActiveXObject(useSpeechPlatform ? "Speech.SpPhoneConverter" : "SAPI.SpPhoneConverter");
29
+ const sapiLanguageCodeHex = sapiVoice.Voice.GetAttribute('Language');
30
+ const sapiLanguageCode = parseInt(sapiLanguageCodeHex, 16);
31
+ sapiPhoneConverter.LanguageId = sapiLanguageCode;
32
+ logger.start("Synthesize with SAPI");
33
+ const sampleRate = 22050;
34
+ const bytesPerSecond = sampleRate * 2;
35
+ const sapiOutputStream = new global.ActiveXObject('SAPI.SpMemoryStream');
36
+ sapiOutputStream.Format.Type = 22; // format code code for SAFT22kHz16BitMono
37
+ sapiVoice.AudioOutputStream = sapiOutputStream;
38
+ const dispatchMessagesInterval = setInterval(() => {
39
+ WinAX.peekAndDispatchMessages();
40
+ }, 50);
41
+ const connectionPoints = WinAX.getConnectionPoints(sapiVoice);
42
+ const connectionPoint = connectionPoints[0];
43
+ const methods = connectionPoint.getMethods();
44
+ const events = [];
45
+ let lastWordEvent = null;
46
+ let lastWordCharPos = -1;
47
+ connectionPoint.advise({
48
+ StartStream: () => {
49
+ },
50
+ Word: (streamId, streamPos, charPos, length) => {
51
+ if (lastWordCharPos == charPos) {
52
+ return;
53
+ }
54
+ const wordText = text.substring(charPos, charPos + length);
55
+ const startTime = streamPos / bytesPerSecond;
56
+ const wordEvent = { type: "word", text: wordText, startTime, endTime: -1, timeline: [] };
57
+ events.push(wordEvent);
58
+ lastWordEvent = wordEvent;
59
+ lastWordCharPos = charPos;
60
+ },
61
+ Phoneme: (streamId, streamPos, duration, nextPhoneId, feature, currentPhoneId) => {
62
+ if (events.length == 0) {
63
+ return;
64
+ }
65
+ const phoneText = sapiPhoneConverter.IdToPhone(currentPhoneId);
66
+ if (phoneText == "," || phoneText == "_") {
67
+ return;
68
+ }
69
+ const startTime = streamPos / bytesPerSecond;
70
+ const endTime = startTime + (duration / 1000);
71
+ events.push({ type: "phone", text: phoneText, startTime, endTime });
72
+ },
73
+ EndStream: (streamId, streamPos) => {
74
+ clearInterval(dispatchMessagesInterval);
75
+ const audioData = Buffer.from(sapiOutputStream.GetData());
76
+ const audioChannels = decodeToChannels(audioData, 1, 16, SampleFormat.PCM);
77
+ WinAX.release(sapiOutputStream);
78
+ WinAX.release(sapiVoice);
79
+ logger.end();
80
+ resolve({ rawAudio: { audioChannels, sampleRate }, timeline: eventsToTimeline(events, audioChannels[0].length / sampleRate) });
81
+ }
82
+ });
83
+ sapiVoice.Speak(text);
84
+ });
85
+ }
86
+ export async function getVoiceList(useSpeechPlatform = false) {
87
+ const { default: WinAX } = await import("winax");
88
+ const sapiVoice = new global.ActiveXObject(useSpeechPlatform ? "Speech.SPVoice" : "SAPI.SPVoice");
89
+ const voiceObjects = sapiVoice.GetVoices();
90
+ const voices = [];
91
+ for (let i = 0; i < voiceObjects.Count; i++) {
92
+ const voiceObject = voiceObjects.Item(i);
93
+ const voiceName = voiceObject.GetDescription();
94
+ const voiceGender = voiceObject.GetAttribute("Gender")?.toLowerCase();
95
+ const sapiLanguageCodeHex = voiceObject.GetAttribute('Language');
96
+ const sapiLanguageCode = parseInt(sapiLanguageCodeHex, 16);
97
+ const languageCodes = await lcidToIsoLanguageCode(sapiLanguageCode);
98
+ if (!languageCodes) {
99
+ throw new Error(`Couldn't translate SAPI language code ${sapiLanguageCode} to ISO, for voice '${voiceName}'`);
100
+ }
101
+ const resultLanguageCodes = [];
102
+ for (const languageCode of languageCodes) {
103
+ if (!resultLanguageCodes.includes(languageCode)) {
104
+ resultLanguageCodes.push(languageCode);
105
+ }
106
+ const shortLanguageCode = getShortLanguageCode(languageCode);
107
+ if (!resultLanguageCodes.includes(shortLanguageCode)) {
108
+ resultLanguageCodes.push(shortLanguageCode);
109
+ }
110
+ }
111
+ voices.push({
112
+ name: voiceName,
113
+ languages: resultLanguageCodes,
114
+ gender: voiceGender || "unknown"
115
+ });
116
+ }
117
+ WinAX.release(sapiVoice);
118
+ return voices;
119
+ }
120
+ export async function AssertSAPIAvailable(testForSpeechPlatform = false) {
121
+ if (process.platform != "win32") {
122
+ throw new Error(`SAPI is not available on your platform. SAPI is a Microsoft Windows technology that is only runs on a Windows OS.`);
123
+ }
124
+ try {
125
+ const { default: WinAX } = await import("winax");
126
+ }
127
+ catch (e) {
128
+ throw new Error(`winax package, which is required for SAPI support, was not found. You can install it by running 'npm install winax -g'.`);
129
+ }
130
+ try {
131
+ const voice = new global.ActiveXObject("SAPI.SPVoice");
132
+ }
133
+ catch (e) {
134
+ throw new Error(`Failed creating a SAPI instance: ${e}`);
135
+ }
136
+ try {
137
+ const voice = new global.ActiveXObject("Speech.SPVoice");
138
+ }
139
+ catch (e) {
140
+ throw new Error(`Failed creating an msspeech instance. Please ensure you installed the Microsoft Speech Platform runtime correctly.`);
141
+ }
142
+ }
143
+ function eventsToTimeline(events, totalDuration) {
144
+ const timeline = [];
145
+ for (const event of events) {
146
+ if (event.type == "word") {
147
+ timeline.push(event);
148
+ }
149
+ else if (event.type == "phone") {
150
+ if (timeline.length == 0) {
151
+ throw new Error("Unexpected: phone event preceded a word event");
152
+ }
153
+ const lastWordEntry = timeline[timeline.length - 1];
154
+ lastWordEntry.endTime = event.endTime;
155
+ const phoneTimeline = lastWordEntry.timeline;
156
+ phoneTimeline.push(event);
157
+ }
158
+ }
159
+ if (timeline.length > 0 && timeline[timeline.length - 1].endTime == -1) {
160
+ timeline[timeline.length - 1].endTime = totalDuration;
161
+ }
162
+ return timeline;
163
+ }
164
+ // References: https://goo.gl/cxlj0b, https://goo.gl/MWbQns, https://goo.gl/sbJxxR
165
+ const lcidToIsoLanguageCodeLookup = {
166
+ 1033: 'en-US',
167
+ 1031: 'de-DE',
168
+ 3081: 'en-AU',
169
+ 4105: 'en-CA',
170
+ 2057: 'en-GB',
171
+ 16393: 'en-IN',
172
+ 1030: 'da-DK',
173
+ 3082: 'es-ES',
174
+ 1035: 'fi-FI',
175
+ 1036: 'fr-FR',
176
+ 1040: 'it-IT',
177
+ 1041: 'ja-JP',
178
+ 1042: 'ko-KR',
179
+ 1043: 'nl-NL',
180
+ 1045: 'pl-PL',
181
+ 2070: 'pt-PT',
182
+ 1049: 'ru-RU',
183
+ 1053: 'sv-SE',
184
+ 2052: 'zh-CN',
185
+ 1025: 'ar-SA',
186
+ 1026: 'bg-BG',
187
+ 1028: 'zh-TW',
188
+ 1050: 'hr-HR',
189
+ 1029: 'cs-CZ',
190
+ 1061: 'et-EE',
191
+ 1032: 'el-GR',
192
+ 1037: 'he-IL',
193
+ 1081: 'hi-IN',
194
+ 1038: 'hu-HU',
195
+ 1057: 'id-ID',
196
+ 1087: 'kk-KZ',
197
+ 1062: 'lv-LV',
198
+ 1063: 'lt-LT',
199
+ 1086: 'ms-MY',
200
+ 1044: 'nb-NO',
201
+ 1046: 'pt-BR',
202
+ 1048: 'ro-RO',
203
+ 2074: 'sr-latn-RS',
204
+ 1051: 'sk-SK',
205
+ 1060: 'sl-SI',
206
+ 1054: 'th-TH',
207
+ 1055: 'tr-TR',
208
+ 1058: 'uk-UA',
209
+ 1066: 'vi-VN',
210
+ };
211
+ //# sourceMappingURL=SapiTTS.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"SapiTTS.js","sourceRoot":"","sources":["../../src/synthesis/SapiTTS.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,gBAAgB,EAAE,MAAM,mCAAmC,CAAA;AAEpE,OAAO,EAAE,YAAY,EAAE,MAAM,wBAAwB,CAAA;AACrD,OAAO,EAAE,oBAAoB,EAAE,qBAAqB,EAAE,MAAM,wBAAwB,CAAA;AACpF,OAAO,EAAE,MAAM,EAAE,MAAM,wBAAwB,CAAA;AAE/C,OAAO,EAAE,WAAW,EAAE,MAAM,2BAA2B,CAAA;AAEvD,MAAM,GAAG,GAAG,WAAW,CAAA;AAEvB,MAAM,UAAU,UAAU,CAAC,IAAY,EAAE,SAAiB,EAAE,IAAI,GAAG,CAAC,EAAE,iBAAiB,GAAG,KAAK;IAC9F,OAAO,IAAI,OAAO,CAA6C,KAAK,EAAE,OAAO,EAAE,MAAM,EAAE,EAAE;QACxF,MAAM,MAAM,GAAG,IAAI,MAAM,EAAE,CAAA;QAC3B,MAAM,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAA;QAEvC,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,OAAO,CAAC,CAAA;QAEhD,MAAM,CAAC,KAAK,CAAC,wBAAwB,CAAC,CAAA;QACtC,MAAM,SAAS,GAAG,IAAI,MAAM,CAAC,aAAa,CAAC,iBAAiB,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,cAAc,CAAC,CAAA;QACjG,SAAS,CAAC,cAAc,GAAG,KAAK,CAAA;QAEhC,MAAM,CAAC,KAAK,CAAC,2CAA2C,CAAC,CAAA;QAEzD,IAAI,SAAS,EAAE;YACd,MAAM,YAAY,GAAG,SAAS,CAAC,SAAS,EAAE,CAAA;YAE1C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,YAAY,CAAC,KAAK,EAAE,CAAC,EAAE,EAAE;gBAC5C,MAAM,WAAW,GAAG,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;gBACxC,MAAM,kBAAkB,GAAG,WAAW,CAAC,cAAc,EAAE,CAAA;gBAEvD,IAAI,kBAAkB,IAAI,SAAS,EAAE;oBACpC,SAAS,CAAC,KAAK,GAAG,WAAW,CAAA;iBAC7B;aACD;SACD;QAED,SAAS,CAAC,IAAI,GAAG,IAAI,CAAA;QAErB,sCAAsC;QACtC,MAAM,kBAAkB,GAAG,IAAI,MAAM,CAAC,aAAa,CAAC,iBAAiB,CAAC,CAAC,CAAC,yBAAyB,CAAC,CAAC,CAAC,uBAAuB,CAAC,CAAA;QAC5H,MAAM,mBAAmB,GAAG,SAAS,CAAC,KAAK,CAAC,YAAY,CAAC,UAAU,CAAC,CAAA;QACpE,MAAM,gBAAgB,GAAG,QAAQ,CAAC,mBAAmB,EAAE,EAAE,CAAC,CAAA;QAE1D,kBAAkB,CAAC,UAAU,GAAG,gBAAgB,CAAA;QAEhD,MAAM,CAAC,KAAK,CAAC,sBAAsB,CAAC,CAAA;QAEpC,MAAM,UAAU,GAAG,KAAK,CAAA;QACxB,MAAM,cAAc,GAAG,UAAU,GAAG,CAAC,CAAA;QAErC,MAAM,gBAAgB,GAAG,IAAI,MAAM,CAAC,aAAa,CAAC,qBAAqB,CAAC,CAAA;QACxE,gBAAgB,CAAC,MAAM,CAAC,IAAI,GAAG,EAAE,CAAA,CAAC,0CAA0C;QAC5E,SAAS,CAAC,iBAAiB,GAAG,gBAAgB,CAAA;QAE9C,MAAM,wBAAwB,GAAG,WAAW,CAAC,GAAG,EAAE;YACjD,KAAK,CAAC,uBAAuB,EAAE,CAAA;QAChC,CAAC,EAAE,EAAE,CAAC,CAAA;QAEN,MAAM,gBAAgB,GAAG,KAAK,CAAC,mBAAmB,CAAC,SAAS,CAAC,CAAA;QAC7D,MAAM,eAAe,GAAG,gBAAgB,CAAC,CAAC,CAAC,CAAA;QAE3C,MAAM,OAAO,GAAG,eAAe,CAAC,UAAU,EAAE,CAAA;QAE5C,MAAM,MAAM,GAAa,EAAE,CAAA;QAC3B,IAAI,aAAa,GAAyB,IAAI,CAAA;QAC9C,IAAI,eAAe,GAAG,CAAC,CAAC,CAAA;QAExB,eAAe,CAAC,MAAM,CAAC;YACtB,WAAW,EAAE,GAAG,EAAE;YAClB,CAAC;YAED,IAAI,EAAE,CAAC,QAAgB,EAAE,SAAiB,EAAE,OAAe,EAAE,MAAc,EAAE,EAAE;gBAC9E,IAAI,eAAe,IAAI,OAAO,EAAE;oBAC/B,OAAM;iBACN;gBAED,MAAM,QAAQ,GAAG,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,OAAO,GAAG,MAAM,CAAC,CAAA;gBAC1D,MAAM,SAAS,GAAG,SAAS,GAAG,cAAc,CAAA;gBAE5C,MAAM,SAAS,GAAG,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,QAAQ,EAAE,SAAS,EAAE,OAAO,EAAE,CAAC,CAAC,EAAE,QAAQ,EAAE,EAAE,EAAmB,CAAA;gBACzG,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,CAAA;gBAEtB,aAAa,GAAG,SAAS,CAAA;gBACzB,eAAe,GAAG,OAAO,CAAA;YAC1B,CAAC;YAED,OAAO,EAAE,CAAC,QAAgB,EAAE,SAAiB,EAAE,QAAgB,EAAE,WAAmB,EAAE,OAAe,EAAE,cAAsB,EAAE,EAAE;gBAChI,IAAI,MAAM,CAAC,MAAM,IAAI,CAAC,EAAE;oBACvB,OAAM;iBACN;gBAED,MAAM,SAAS,GAAG,kBAAkB,CAAC,SAAS,CAAC,cAAc,CAAC,CAAA;gBAE9D,IAAI,SAAS,IAAI,GAAG,IAAI,SAAS,IAAI,GAAG,EAAE;oBACzC,OAAM;iBACN;gBAED,MAAM,SAAS,GAAG,SAAS,GAAG,cAAc,CAAA;gBAC5C,MAAM,OAAO,GAAG,SAAS,GAAG,CAAC,QAAQ,GAAG,IAAI,CAAC,CAAA;gBAE7C,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,OAAO,EAAE,CAAC,CAAA;YACpE,CAAC;YAED,SAAS,EAAE,CAAC,QAAgB,EAAE,SAAiB,EAAE,EAAE;gBAClD,aAAa,CAAC,wBAAwB,CAAC,CAAA;gBAEvC,MAAM,SAAS,GAAG,MAAM,CAAC,IAAI,CAAC,gBAAgB,CAAC,OAAO,EAAE,CAAC,CAAA;gBACzD,MAAM,aAAa,GAAG,gBAAgB,CAAC,SAAS,EAAE,CAAC,EAAE,EAAE,EAAE,YAAY,CAAC,GAAG,CAAC,CAAA;gBAE1E,KAAK,CAAC,OAAO,CAAC,gBAAgB,CAAC,CAAA;gBAC/B,KAAK,CAAC,OAAO,CAAC,SAAS,CAAC,CAAA;gBAExB,MAAM,CAAC,GAAG,EAAE,CAAA;gBAEZ,OAAO,CAAC,EAAE,QAAQ,EAAE,EAAE,aAAa,EAAE,UAAU,EAAE,EAAE,QAAQ,EAAE,gBAAgB,CAAC,MAAM,EAAE,aAAa,CAAC,CAAC,CAAC,CAAC,MAAM,GAAG,UAAU,CAAC,EAAE,CAAC,CAAA;YAC/H,CAAC;SACD,CAAC,CAAA;QAEF,SAAS,CAAC,KAAK,CAAC,IAAI,CAAC,CAAA;IACtB,CAAC,CAAC,CAAA;AACH,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,YAAY,CAAC,iBAAiB,GAAG,KAAK;IAC3D,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,OAAO,CAAC,CAAA;IAEhD,MAAM,SAAS,GAAG,IAAI,MAAM,CAAC,aAAa,CAAC,iBAAiB,CAAC,CAAC,CAAC,gBAAgB,CAAC,CAAC,CAAC,cAAc,CAAC,CAAA;IAEjG,MAAM,YAAY,GAAG,SAAS,CAAC,SAAS,EAAE,CAAA;IAE1C,MAAM,MAAM,GAAqB,EAAE,CAAA;IAEnC,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,YAAY,CAAC,KAAK,EAAE,CAAC,EAAE,EAAE;QAC5C,MAAM,WAAW,GAAG,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,CAAA;QACxC,MAAM,SAAS,GAAG,WAAW,CAAC,cAAc,EAAE,CAAA;QAC9C,MAAM,WAAW,GAAG,WAAW,CAAC,YAAY,CAAC,QAAQ,CAAC,EAAE,WAAW,EAAE,CAAA;QAErE,MAAM,mBAAmB,GAAG,WAAW,CAAC,YAAY,CAAC,UAAU,CAAC,CAAA;QAChE,MAAM,gBAAgB,GAAG,QAAQ,CAAC,mBAAmB,EAAE,EAAE,CAAC,CAAA;QAE1D,MAAM,aAAa,GAAG,MAAM,qBAAqB,CAAC,gBAAgB,CAAC,CAAA;QAEnE,IAAI,CAAC,aAAa,EAAE;YACnB,MAAM,IAAI,KAAK,CAAC,yCAAyC,gBAAgB,uBAAuB,SAAS,GAAG,CAAC,CAAA;SAC7G;QAED,MAAM,mBAAmB,GAAa,EAAE,CAAA;QAExC,KAAK,MAAM,YAAY,IAAI,aAAa,EAAE;YACzC,IAAI,CAAC,mBAAmB,CAAC,QAAQ,CAAC,YAAY,CAAC,EAAE;gBAChD,mBAAmB,CAAC,IAAI,CAAC,YAAY,CAAC,CAAA;aACtC;YAED,MAAM,iBAAiB,GAAG,oBAAoB,CAAC,YAAY,CAAC,CAAA;YAC5D,IAAI,CAAC,mBAAmB,CAAC,QAAQ,CAAC,iBAAiB,CAAC,EAAE;gBACrD,mBAAmB,CAAC,IAAI,CAAC,iBAAiB,CAAC,CAAA;aAC3C;SACD;QAED,MAAM,CAAC,IAAI,CAAC;YACX,IAAI,EAAE,SAAS;YACf,SAAS,EAAE,mBAAmB;YAC9B,MAAM,EAAE,WAAW,IAAI,SAAS;SAChC,CAAC,CAAA;KACF;IAED,KAAK,CAAC,OAAO,CAAC,SAAS,CAAC,CAAA;IAExB,OAAO,MAAM,CAAA;AACd,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,mBAAmB,CAAC,qBAAqB,GAAG,KAAK;IACtE,IAAI,OAAO,CAAC,QAAQ,IAAI,OAAO,EAAE;QAChC,MAAM,IAAI,KAAK,CAAC,mHAAmH,CAAC,CAAA;KACpI;IAED,IAAI;QACH,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,OAAO,CAAC,CAAA;KAChD;IAAC,OAAO,CAAC,EAAE;QACX,MAAM,IAAI,KAAK,CAAC,yHAAyH,CAAC,CAAA;KAC1I;IAED,IAAI;QACH,MAAM,KAAK,GAAG,IAAI,MAAM,CAAC,aAAa,CAAC,cAAc,CAAC,CAAA;KACtD;IAAC,OAAO,CAAC,EAAE;QACX,MAAM,IAAI,KAAK,CAAC,oCAAoC,CAAC,EAAE,CAAC,CAAA;KACxD;IAED,IAAI;QACH,MAAM,KAAK,GAAG,IAAI,MAAM,CAAC,aAAa,CAAC,gBAAgB,CAAC,CAAA;KACxD;IAAC,OAAM,CAAC,EAAE;QACV,MAAM,IAAI,KAAK,CAAC,oHAAoH,CAAC,CAAA;KACrI;AACF,CAAC;AAED,SAAS,gBAAgB,CAAC,MAAgB,EAAE,aAAqB;IAChE,MAAM,QAAQ,GAAa,EAAE,CAAA;IAE7B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE;QAC3B,IAAI,KAAK,CAAC,IAAI,IAAI,MAAM,EAAE;YACzB,QAAQ,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;SACpB;aAAM,IAAI,KAAK,CAAC,IAAI,IAAI,OAAO,EAAE;YACjC,IAAI,QAAQ,CAAC,MAAM,IAAI,CAAC,EAAE;gBACzB,MAAM,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAA;aAChE;YAED,MAAM,aAAa,GAAG,QAAQ,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAA;YAEnD,aAAa,CAAC,OAAO,GAAG,KAAK,CAAC,OAAO,CAAA;YAErC,MAAM,aAAa,GAAG,aAAa,CAAC,QAAS,CAAA;YAE7C,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;SACzB;KACD;IAED,IAAI,QAAQ,CAAC,MAAM,GAAG,CAAC,IAAI,QAAQ,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,OAAO,IAAI,CAAC,CAAC,EAAE;QACvE,QAAQ,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,OAAO,GAAG,aAAa,CAAA;KACrD;IAED,OAAO,QAAQ,CAAA;AAChB,CAAC;AAED,kFAAkF;AAClF,MAAM,2BAA2B,GAA4B;IAC5D,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,KAAK,EAAE,OAAO;IACd,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IAEb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,YAAY;IAClB,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;IACb,IAAI,EAAE,OAAO;CACb,CAAA"}
@@ -0,0 +1,12 @@
1
+ /// <reference types="node" resolution-mode="require"/>
2
+ import { SynthesisVoice } from "../api/API.js";
3
+ import { Timeline } from "../utilities/Timeline.js";
4
+ export declare function synthesizeLongText(text: string, voice: string, languageCode: string, sentenceEndPause?: number, segmentEndPause?: number): Promise<{
5
+ rawAudio: {
6
+ audioChannels: Float32Array[];
7
+ sampleRate: number;
8
+ };
9
+ timeline: Timeline;
10
+ }>;
11
+ export declare function synthesizeFragment(text: string, voice: string): Promise<Buffer>;
12
+ export declare const voiceList: SynthesisVoice[];
@@ -0,0 +1,88 @@
1
+ import { request } from "gaxios";
2
+ import { trimAudioEnd } from "../audio/AudioUtilities.js";
3
+ import * as FFMpegTranscoder from "../codecs/FFMpegTranscoder.js";
4
+ import { Phrase, splitToFragments } from "../nlp/Segmentation.js";
5
+ import { Logger } from "../utilities/Logger.js";
6
+ import { concatFloat32Arrays, logToStderr } from "../utilities/Utilities.js";
7
+ const log = logToStderr;
8
+ const maxTextLengthPerRequest = 200;
9
+ export async function synthesizeLongText(text, voice, languageCode, sentenceEndPause = 0.75, segmentEndPause = 1.0) {
10
+ if (text.length == 0) {
11
+ throw new Error("Text is empty");
12
+ }
13
+ const logger = new Logger();
14
+ logger.start("Prepare and split text");
15
+ const fragments = await splitToFragments(text, maxTextLengthPerRequest, languageCode);
16
+ const audioFragments = [];
17
+ let fragmentsSampleRate = 0;
18
+ const timeline = [];
19
+ for (let i = 0; i < fragments.length; i++) {
20
+ const fragment = fragments[i];
21
+ logger.start(`Request synthesis for text fragment ${i + 1}/${fragments.length} from Streamslabs Polly`);
22
+ const fragmentMp3Stream = await synthesizeFragment(fragment.text, voice);
23
+ if (fragmentMp3Stream.length == 0) {
24
+ continue;
25
+ }
26
+ const rawAudio = await FFMpegTranscoder.decodeToChannels(fragmentMp3Stream, 24000, 1);
27
+ fragmentsSampleRate = rawAudio.sampleRate;
28
+ let targetEndingSilenceTime;
29
+ if (fragment.lastSegment?.isSentenceFinalizer) {
30
+ targetEndingSilenceTime = sentenceEndPause;
31
+ }
32
+ else if (fragment.lastSegment instanceof Phrase) {
33
+ targetEndingSilenceTime = segmentEndPause;
34
+ }
35
+ else {
36
+ targetEndingSilenceTime = 0;
37
+ }
38
+ const trimmedAudio = trimAudioEnd(rawAudio.audioChannels[0], targetEndingSilenceTime * rawAudio.sampleRate);
39
+ audioFragments.push(trimmedAudio);
40
+ const startTime = timeline.length == 0 ? 0 : timeline[timeline.length - 1].endTime;
41
+ const endTime = startTime + (trimmedAudio.length / fragmentsSampleRate);
42
+ timeline.push({
43
+ type: "segment",
44
+ text: fragment.text,
45
+ startTime,
46
+ endTime
47
+ });
48
+ }
49
+ const rawAudio = { audioChannels: [concatFloat32Arrays(audioFragments)], sampleRate: fragmentsSampleRate };
50
+ logger.end();
51
+ return {
52
+ rawAudio,
53
+ timeline
54
+ };
55
+ }
56
+ export async function synthesizeFragment(text, voice) {
57
+ const response = await request({
58
+ url: `https://streamlabs.com/polly/speak`,
59
+ method: "POST",
60
+ data: {
61
+ voice,
62
+ text,
63
+ },
64
+ responseType: "json"
65
+ });
66
+ const responseObject = response.data;
67
+ const audioUrl = responseObject.speak_url;
68
+ const audioUrlResponse = await request({ url: audioUrl, responseType: "arraybuffer" });
69
+ return Buffer.from(audioUrlResponse.data);
70
+ }
71
+ export const voiceList = [
72
+ { name: "Brian", languages: ["en-GB", "en"], gender: "male" },
73
+ { name: "Emma", languages: ["en-GB", "en"], gender: "female" },
74
+ { name: "Russell", languages: ["en-AU", "en"], gender: "male" },
75
+ { name: "Joey", languages: ["en-US", "en"], gender: "male" },
76
+ { name: "Matthew", languages: ["en-US", "en"], gender: "male" },
77
+ { name: "Joanna", languages: ["en-US", "en"], gender: "female" },
78
+ { name: "Kimberly", languages: ["en-US", "en"], gender: "female" },
79
+ { name: "Amy", languages: ["en-GB", "en"], gender: "female" },
80
+ { name: "Geraint", languages: ["en-GB-WLS", "en-GB", "en"], gender: "male" },
81
+ { name: "Nicole", languages: ["en-AU", "en"], gender: "female" },
82
+ { name: "Justin", languages: ["en-US", "en"], gender: "male" },
83
+ { name: "Ivy", languages: ["en-US", "en"], gender: "female" },
84
+ { name: "Kendra", languages: ["en-US", "en"], gender: "female" },
85
+ { name: "Salli", languages: ["en-US", "en"], gender: "female" },
86
+ { name: "Raveena", languages: ["en-IN", "en"], gender: "female" },
87
+ ];
88
+ //# sourceMappingURL=StreamlabsPollyTTS.js.map
@@ -0,0 +1 @@
1
+ {"version":3,"file":"StreamlabsPollyTTS.js","sourceRoot":"","sources":["../../src/synthesis/StreamlabsPollyTTS.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,OAAO,EAAE,MAAM,QAAQ,CAAA;AAEhC,OAAO,EAAE,YAAY,EAAE,MAAM,4BAA4B,CAAA;AACzD,OAAO,KAAK,gBAAgB,MAAM,+BAA+B,CAAA;AACjE,OAAO,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM,wBAAwB,CAAA;AACjE,OAAO,EAAE,MAAM,EAAE,MAAM,wBAAwB,CAAA;AAC/C,OAAO,EAAE,mBAAmB,EAAE,WAAW,EAAE,MAAM,2BAA2B,CAAA;AAG5E,MAAM,GAAG,GAAG,WAAW,CAAA;AAEvB,MAAM,uBAAuB,GAAG,GAAG,CAAA;AAEnC,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,IAAY,EAAE,KAAa,EAAE,YAAoB,EAAE,gBAAgB,GAAG,IAAI,EAAE,eAAe,GAAG,GAAG;IACzI,IAAI,IAAI,CAAC,MAAM,IAAI,CAAC,EAAE;QACrB,MAAM,IAAI,KAAK,CAAC,eAAe,CAAC,CAAA;KAChC;IAED,MAAM,MAAM,GAAG,IAAI,MAAM,EAAE,CAAA;IAC3B,MAAM,CAAC,KAAK,CAAC,wBAAwB,CAAC,CAAA;IAEtC,MAAM,SAAS,GAAG,MAAM,gBAAgB,CAAC,IAAI,EAAE,uBAAuB,EAAE,YAAY,CAAC,CAAA;IAErF,MAAM,cAAc,GAAmB,EAAE,CAAA;IACzC,IAAI,mBAAmB,GAAG,CAAC,CAAA;IAE3B,MAAM,QAAQ,GAAa,EAAE,CAAA;IAE7B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,SAAS,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE;QAC1C,MAAM,QAAQ,GAAG,SAAS,CAAC,CAAC,CAAC,CAAA;QAE7B,MAAM,CAAC,KAAK,CAAC,uCAAuC,CAAC,GAAG,CAAC,IAAI,SAAS,CAAC,MAAM,yBAAyB,CAAC,CAAA;QACvG,MAAM,iBAAiB,GAAG,MAAM,kBAAkB,CAAC,QAAQ,CAAC,IAAI,EAAE,KAAK,CAAC,CAAA;QAExE,IAAI,iBAAiB,CAAC,MAAM,IAAI,CAAC,EAAE;YAClC,SAAQ;SACR;QAED,MAAM,QAAQ,GAAG,MAAM,gBAAgB,CAAC,gBAAgB,CAAC,iBAAiB,EAAE,KAAK,EAAE,CAAC,CAAC,CAAA;QACrF,mBAAmB,GAAG,QAAQ,CAAC,UAAU,CAAA;QAEzC,IAAI,uBAA+B,CAAA;QAEnC,IAAI,QAAQ,CAAC,WAAW,EAAE,mBAAmB,EAAE;YAC9C,uBAAuB,GAAG,gBAAgB,CAAA;SAC1C;aAAM,IAAI,QAAQ,CAAC,WAAW,YAAY,MAAM,EAAE;YAClD,uBAAuB,GAAG,eAAe,CAAA;SACzC;aAAM;YACN,uBAAuB,GAAG,CAAC,CAAA;SAC3B;QAED,MAAM,YAAY,GAAG,YAAY,CAAC,QAAQ,CAAC,aAAa,CAAC,CAAC,CAAC,EAAE,uBAAuB,GAAG,QAAQ,CAAC,UAAU,CAAC,CAAA;QAE3G,cAAc,CAAC,IAAI,CAAC,YAAY,CAAC,CAAA;QAEjC,MAAM,SAAS,GAAG,QAAQ,CAAC,MAAM,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,OAAO,CAAA;QAClF,MAAM,OAAO,GAAG,SAAS,GAAG,CAAC,YAAY,CAAC,MAAM,GAAG,mBAAmB,CAAC,CAAA;QAEvE,QAAQ,CAAC,IAAI,CAAC;YACb,IAAI,EAAE,SAAS;YACf,IAAI,EAAE,QAAQ,CAAC,IAAI;YACnB,SAAS;YACT,OAAO;SACP,CAAC,CAAA;KACF;IAED,MAAM,QAAQ,GAAG,EAAE,aAAa,EAAE,CAAC,mBAAmB,CAAC,cAAc,CAAC,CAAC,EAAE,UAAU,EAAE,mBAAmB,EAAE,CAAA;IAE1G,MAAM,CAAC,GAAG,EAAE,CAAA;IAEZ,OAAO;QACN,QAAQ;QACR,QAAQ;KACR,CAAA;AACF,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,IAAY,EAAE,KAAa;IACnE,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAM;QACnC,GAAG,EAAE,oCAAoC;QAEzC,MAAM,EAAE,MAAM;QAEd,IAAI,EAAE;YACL,KAAK;YACL,IAAI;SACJ;QAED,YAAY,EAAE,MAAM;KACpB,CAAC,CAAA;IAEF,MAAM,cAAc,GAAG,QAAQ,CAAC,IAAI,CAAA;IACpC,MAAM,QAAQ,GAAG,cAAc,CAAC,SAAS,CAAA;IACzC,MAAM,gBAAgB,GAAG,MAAM,OAAO,CAAc,EAAE,GAAG,EAAE,QAAQ,EAAE,YAAY,EAAE,aAAa,EAAE,CAAC,CAAA;IAEnG,OAAO,MAAM,CAAC,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAA;AAC1C,CAAC;AAED,MAAM,CAAC,MAAM,SAAS,GAAqB;IAC1C,EAAE,IAAI,EAAE,OAAO,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE;IAC7D,EAAE,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAC9D,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE;IAC/D,EAAE,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE;IAC5D,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE;IAC/D,EAAE,IAAI,EAAE,QAAQ,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAChE,EAAE,IAAI,EAAE,UAAU,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAClE,EAAE,IAAI,EAAE,KAAK,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAC7D,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,CAAC,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE;IAC5E,EAAE,IAAI,EAAE,QAAQ,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAChE,EAAE,IAAI,EAAE,QAAQ,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,MAAM,EAAE;IAC9D,EAAE,IAAI,EAAE,KAAK,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAC7D,EAAE,IAAI,EAAE,QAAQ,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAChE,EAAE,IAAI,EAAE,OAAO,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;IAC/D,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,MAAM,EAAE,QAAQ,EAAE;CACjE,CAAA"}
@@ -0,0 +1,11 @@
1
+ import { SynthesisVoice } from "../api/API.js";
2
+ import { RawAudio } from "../audio/AudioUtilities.js";
3
+ export declare function synthesize(text: string, textAnalysisFilePath: string, signalGenerationFilePath: string, postprocessOutput?: boolean): Promise<{
4
+ rawAudio: RawAudio;
5
+ }>;
6
+ export declare function getInstance(): Promise<any>;
7
+ export declare function getResourceFilenamesForLanguage(language: string): {
8
+ textAnalysisFilename: string;
9
+ signalGenerationFilename: string;
10
+ };
11
+ export declare const voiceList: SynthesisVoice[];