0.0.49

update to azure speech sdk 1.36.0
0.0.48
2026-10-03 23:33:59 +00:00 · 2024-04-07 12:15:02 -04:00 · 2024-04-07 12:14:55 -04:00 · 2024-04-04 15:52:40 -04:00 · 2024-04-04 15:52:10 -04:00 · 2024-04-04 18:58:38 +07:00
4 changed files with 207 additions and 70 deletions
@@ -141,12 +141,12 @@ async function synthAudio(client, logger, stats, { account_sid,
  let filePath;
  if (['nuance', 'nvidia'].includes(vendor) ||
    (
-      process.env.JAMBONES_TTS_TRIM_SILENCE &&
+      (process.env.JAMBONES_TTS_TRIM_SILENCE || !process.env.JAMBONES_DISABLE_TTS_STREAMING) &&
      ['microsoft', 'azure'].includes(vendor)
    ) ||
    (
      !process.env.JAMBONES_DISABLE_TTS_STREAMING &&
-      vendor === 'elevenlabs'
+      ['elevenlabs', 'deepgram'].includes(vendor)
    )
  ) {
    filePath = `${TMP_FOLDER}/${key.replace('tts:', `tts-${salt || ''}`)}.r8`;
@@ -183,7 +183,9 @@ async function synthAudio(client, logger, stats, { account_sid,
      case 'azure':
      case 'microsoft':
        vendorLabel = 'microsoft';
-        audioBuffer = await synthMicrosoft(logger, {credentials, stats, language, voice, text, deploymentId, filePath});
+        audioBuffer = await synthMicrosoft(logger, {credentials, stats, language, voice, text, deploymentId,
+          filePath, renderForCaching, disableTtsStreaming});
+        if (audioBuffer?.filePath) return audioBuffer;
        break;
      case 'nuance':
        model = model || 'enhanced';
@@ -210,7 +212,9 @@ async function synthAudio(client, logger, stats, { account_sid,
        if (audioBuffer?.filePath) return audioBuffer;
        break;
      case 'deepgram':
-        audioBuffer = await synthDeepgram(logger, {credentials, stats, model, text});
+        audioBuffer = await synthDeepgram(logger, {credentials, stats, model, text,
+          renderForCaching, disableTtsStreaming});
+        if (audioBuffer?.filePath) return audioBuffer;
        break;
      case vendor.startsWith('custom') ? vendor : 'cant_match_value':
        ({ audioBuffer, filePath } = await synthCustomVendor(logger,
@@ -381,10 +385,47 @@ const synthMicrosoft = async(logger, {
  language,
  voice,
  text,
-  filePath
+  filePath,
+  renderForCaching,
+  disableTtsStreaming
 }) => {
  try {
    const {api_key: apiKey, region, use_custom_tts, custom_tts_endpoint, custom_tts_endpoint_url} = credentials;
+    // let clean up the text
+    let content = text;
+    if (use_custom_tts && !content.startsWith('<speak')) {
+      /**
+       * Note: it seems that to use custom voice ssml is required with the voice attribute
+       * Otherwise sending plain text we get "Voice does not match"
+       */
+      content = `<speak>${text}</speak>`;
+    }
+
+    if (content.startsWith('<speak>')) {
+      /* microsoft enforces some properties and uses voice xml element so if the user did not supply do it for them */
+      const words = content.slice(7, -8).trim().replace(/(\r\n|\n|\r)/gm, ' ');
+      // eslint-disable-next-line max-len
+      content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}">${words}</voice></speak>`;
+      logger.info({content}, 'synthMicrosoft');
+    }
+    if (!process.env.JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+      let params = '';
+      params += `{api_key=${apiKey}`;
+      params += `,language=${language}`;
+      params += ',vendor=microsoft';
+      params += `,voice=${voice}`;
+      params += ',write_cache_file=1';
+      if (region) params += `,region=${region}`;
+      if (custom_tts_endpoint) params += `,endpointId=${custom_tts_endpoint}`;
+      if (process.env.JAMBONES_HTTP_PROXY_IP) params += `,http_proxy_ip=${process.env.JAMBONES_HTTP_PROXY_IP}`;
+      if (process.env.JAMBONES_HTTP_PROXY_PORT) params += `,http_proxy_port=${process.env.JAMBONES_HTTP_PROXY_PORT}`;
+      params += '}';
+      return {
+        filePath: `say:${params}${content.replace(/\n/g, ' ')}`,
+        servedFromCache: false,
+        rtt: 0
+      };
+    }
    if (use_custom_tts && custom_tts_endpoint_url) {
      return await _synthOnPremMicrosoft(logger, {
        credentials,
@@ -396,20 +437,12 @@ const synthMicrosoft = async(logger, {
      });
    }
    const trimSilence = filePath.endsWith('.r8');
-    let content = text;
    const speechConfig = SpeechConfig.fromSubscription(apiKey, region);
    speechConfig.speechSynthesisLanguage = language;
    speechConfig.speechSynthesisVoiceName = voice;
    if (use_custom_tts && custom_tts_endpoint) {
      speechConfig.endpointId = custom_tts_endpoint;
    }
-    if (use_custom_tts && !content.startsWith('<speak')) {
-      /**
-       * Note: it seems that to use custom voice ssml is required with the voice attribute
-       * Otherwise sending plain text we get "Voice does not match"
-       */
-      content = `<speak>${text}</speak>`;
-    }
    speechConfig.speechSynthesisOutputFormat = trimSilence ?
      SpeechSynthesisOutputFormat.Raw8Khz16BitMonoPcm :
      SpeechSynthesisOutputFormat.Audio16Khz32KBitRateMonoMp3;
@@ -421,14 +454,6 @@ const synthMicrosoft = async(logger, {
    }
    const synthesizer = new SpeechSynthesizer(speechConfig);

-    if (content.startsWith('<speak>')) {
-      /* microsoft enforces some properties and uses voice xml element so if the user did not supply do it for them */
-      const words = content.slice(7, -8).trim().replace(/(\r\n|\n|\r)/gm, ' ');
-      // eslint-disable-next-line max-len
-      content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="${language}"><voice name="${voice}">${words}</voice></speak>`;
-      logger.info({content}, 'synthMicrosoft');
-    }
-
    return new Promise((resolve, reject) => {
      const speakAsync = content.startsWith('<speak') ?
        synthesizer.speakSsmlAsync.bind(synthesizer) :
@@ -614,6 +639,8 @@ const synthElevenlabs = async(logger, {
  if (!process.env.JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
    let params = '';
    params += `{api_key=${api_key}`;
+    params += ',vendor=elevenlabs';
+    params += `,voice=${voice}`;
    params += `,model_id=${model_id}`;
    params += `,optimize_streaming_latency=${opts.optimize_streaming_latency || 2}`;
    params += ',write_cache_file=1';
@@ -662,6 +689,7 @@ const synthWhisper = async(logger, {credentials, stats, voice, text, renderForCa
    let params = '';
    params += `{api_key=${api_key}`;
    params += `,model_id=${model_id}`;
+    params += ',vendor=whisper';
    params += `,voice=${voice}`;
    params += ',write_cache_file=1';
    if (speed) params += `,speed=${speed}`;
@@ -694,8 +722,22 @@ const synthWhisper = async(logger, {credentials, stats, voice, text, renderForCa
  }
 };

-const synthDeepgram = async(logger, {credentials, stats, model, text}) => {
+const synthDeepgram = async(logger, {credentials, stats, model, text, renderForCaching, disableTtsStreaming}) => {
  const {api_key} = credentials;
+  if (!process.env.JAMBONES_DISABLE_TTS_STREAMING && !renderForCaching && !disableTtsStreaming) {
+    let params = '';
+    params += `{api_key=${api_key}`;
+    params += ',vendor=deepgram';
+    params += `,voice=${model}`;
+    params += ',write_cache_file=1';
+    params += '}';
+
+    return {
+      filePath: `say:${params}${text.replace(/\n/g, ' ')}`,
+      servedFromCache: false,
+      rtt: 0
+    };
+  }
  try {
    const post = bent('https://api.deepgram.com', 'POST', 'buffer', {
      'Authorization': `Token ${api_key}`,
@@ -1,12 +1,12 @@
 {
  "name": "@jambonz/speech-utils",
-  "version": "0.0.43",
+  "version": "0.0.49",
  "lockfileVersion": 2,
  "requires": true,
  "packages": {
    "": {
      "name": "@jambonz/speech-utils",
-      "version": "0.0.43",
+      "version": "0.0.49",
      "license": "MIT",
      "dependencies": {
        "@aws-sdk/client-polly": "^3.496.0",
@@ -19,7 +19,7 @@
        "form-urlencoded": "^6.1.4",
        "google-protobuf": "^3.21.2",
        "ibm-watson": "^8.0.0",
-        "microsoft-cognitiveservices-speech-sdk": "1.34.0",
+        "microsoft-cognitiveservices-speech-sdk": "1.36.0",
        "openai": "^4.25.0",
        "undici": "^6.4.0"
      },
@@ -1124,14 +1124,6 @@
        "node": "^12.22.0 || ^14.17.0 || >=16.0.0"
      }
    },
-    "node_modules/@fastify/busboy": {
-      "version": "2.1.0",
-      "resolved": "https://registry.npmjs.org/@fastify/busboy/-/busboy-2.1.0.tgz",
-      "integrity": "sha512-+KpH+QxZU7O4675t3mnkQKcZZg56u+K/Ct2K+N2AZYNVK8kyeo/bI18tI8aPm3tvNNRyTWfj6s5tnGNlcbQRsA==",
-      "engines": {
-        "node": ">=14"
-      }
-    },
    "node_modules/@google-cloud/text-to-speech": {
      "version": "5.0.2",
      "resolved": "https://registry.npmjs.org/@google-cloud/text-to-speech/-/text-to-speech-5.0.2.tgz",
@@ -3125,13 +3117,14 @@
      }
    },
    "node_modules/es5-ext": {
-      "version": "0.10.62",
-      "resolved": "https://registry.npmjs.org/es5-ext/-/es5-ext-0.10.62.tgz",
-      "integrity": "sha512-BHLqn0klhEpnOKSrzn/Xsz2UIW8j+cGmo9JLzr8BiUapV8hPL9+FliFqjwr9ngW7jWdnxv6eO+/LqyhJVqgrjA==",
+      "version": "0.10.64",
+      "resolved": "https://registry.npmjs.org/es5-ext/-/es5-ext-0.10.64.tgz",
+      "integrity": "sha512-p2snDhiLaXe6dahss1LddxqEm+SkuDvV8dnIQG0MWjyHpcMNfXKPE+/Cc0y+PhxJX3A4xGNeFCj5oc0BUh6deg==",
      "hasInstallScript": true,
      "dependencies": {
        "es6-iterator": "^2.0.3",
        "es6-symbol": "^3.1.3",
+        "esniff": "^2.0.1",
        "next-tick": "^1.1.0"
      },
      "engines": {
@@ -3278,6 +3271,25 @@
        "url": "https://opencollective.com/eslint"
      }
    },
+    "node_modules/esniff": {
+      "version": "2.0.1",
+      "resolved": "https://registry.npmjs.org/esniff/-/esniff-2.0.1.tgz",
+      "integrity": "sha512-kTUIGKQ/mDPFoJ0oVfcmyJn4iBDRptjNVIzwIFR7tqWXdVI9xfA2RMwY/gbSpJG3lkdWNEjLap/NqVHZiJsdfg==",
+      "dependencies": {
+        "d": "^1.0.1",
+        "es5-ext": "^0.10.62",
+        "event-emitter": "^0.3.5",
+        "type": "^2.7.2"
+      },
+      "engines": {
+        "node": ">=0.10"
+      }
+    },
+    "node_modules/esniff/node_modules/type": {
+      "version": "2.7.2",
+      "resolved": "https://registry.npmjs.org/type/-/type-2.7.2.tgz",
+      "integrity": "sha512-dzlvlNlt6AXU7EBSfpAscydQ7gXB+pPGsPnfJnZpiNJBDj7IaJzQlBZYGdEi4R9HmPdBv2XmWJ6YUtoTa7lmCw=="
+    },
    "node_modules/espree": {
      "version": "9.6.1",
      "resolved": "https://registry.npmjs.org/espree/-/espree-9.6.1.tgz",
@@ -3350,6 +3362,15 @@
        "node": ">=0.10.0"
      }
    },
+    "node_modules/event-emitter": {
+      "version": "0.3.5",
+      "resolved": "https://registry.npmjs.org/event-emitter/-/event-emitter-0.3.5.tgz",
+      "integrity": "sha512-D9rRn9y7kLPnJ+hMq7S/nhvoKwwvVJahBi2BPmx3bvbsEdK3W9ii8cBSGjP+72/LnM4n6fo3+dkCX5FeTQruXA==",
+      "dependencies": {
+        "d": "1",
+        "es5-ext": "~0.10.14"
+      }
+    },
    "node_modules/event-target-shim": {
      "version": "5.0.1",
      "resolved": "https://registry.npmjs.org/event-target-shim/-/event-target-shim-5.0.1.tgz",
@@ -3551,9 +3572,9 @@
      "dev": true
    },
    "node_modules/follow-redirects": {
-      "version": "1.15.5",
-      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.5.tgz",
-      "integrity": "sha512-vSFWUON1B+yAw1VN4xMfxgn5fTUiaOzAJCKBwIIgT/+7CuGy9+r+5gITvP62j3RmaD5Ph65UaERdOSRGUzZtgw==",
+      "version": "1.15.6",
+      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.6.tgz",
+      "integrity": "sha512-wWN62YITEaOpSK584EZXJafH1AGpO8RVgElfkuXbTOrPX4fIfOyEpW/CsiNd8JdYrAoOvafRTOEnvsO++qCqFA==",
      "funding": [
        {
          "type": "individual",
@@ -5084,9 +5105,9 @@
      }
    },
    "node_modules/microsoft-cognitiveservices-speech-sdk": {
-      "version": "1.34.0",
-      "resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.34.0.tgz",
-      "integrity": "sha512-WAR0YqouRzVux2kI+f5wTPC6NyJgIVC1g65d79dJ9I32WPJs2kK+eb/BMB6mhSdCjackO5FsrW7JLaQ/vB1heQ==",
+      "version": "1.36.0",
+      "resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.36.0.tgz",
+      "integrity": "sha512-wPxuEXgjLdqMMIrdBtl8jquGahLV19LQE0ie8MI/PcBcNLG5buVzwS2rQEyHMsRGx+C/4OdBo1ROdNIUzCm4Lg==",
      "dependencies": {
        "@types/webrtc": "^0.0.37",
        "agent-base": "^6.0.1",
@@ -6851,12 +6872,9 @@
      }
    },
    "node_modules/undici": {
-      "version": "6.4.0",
-      "resolved": "https://registry.npmjs.org/undici/-/undici-6.4.0.tgz",
-      "integrity": "sha512-wYaKgftNqf6Je7JQ51YzkEkEevzOgM7at5JytKO7BjaURQpERW8edQSMrr2xb+Yv4U8Yg47J24+lc9+NbeXMFA==",
-      "dependencies": {
-        "@fastify/busboy": "^2.0.0"
-      },
+      "version": "6.11.1",
+      "resolved": "https://registry.npmjs.org/undici/-/undici-6.11.1.tgz",
+      "integrity": "sha512-KyhzaLJnV1qa3BSHdj4AZ2ndqI0QWPxYzaIOio0WzcEJB9gvuysprJSLtpvc2D9mhR9jPDUk7xlJlZbH2KR5iw==",
      "engines": {
        "node": ">=18.0"
      }
@@ -8092,11 +8110,6 @@
      "integrity": "sha512-gMsVel9D7f2HLkBma9VbtzZRehRogVRfbr++f06nL2vnCGCNlzOD+/MUov/F4p8myyAHspEhVobgjpX64q5m6A==",
      "dev": true
    },
-    "@fastify/busboy": {
-      "version": "2.1.0",
-      "resolved": "https://registry.npmjs.org/@fastify/busboy/-/busboy-2.1.0.tgz",
-      "integrity": "sha512-+KpH+QxZU7O4675t3mnkQKcZZg56u+K/Ct2K+N2AZYNVK8kyeo/bI18tI8aPm3tvNNRyTWfj6s5tnGNlcbQRsA=="
-    },
    "@google-cloud/text-to-speech": {
      "version": "5.0.2",
      "resolved": "https://registry.npmjs.org/@google-cloud/text-to-speech/-/text-to-speech-5.0.2.tgz",
@@ -9652,12 +9665,13 @@
      }
    },
    "es5-ext": {
-      "version": "0.10.62",
-      "resolved": "https://registry.npmjs.org/es5-ext/-/es5-ext-0.10.62.tgz",
-      "integrity": "sha512-BHLqn0klhEpnOKSrzn/Xsz2UIW8j+cGmo9JLzr8BiUapV8hPL9+FliFqjwr9ngW7jWdnxv6eO+/LqyhJVqgrjA==",
+      "version": "0.10.64",
+      "resolved": "https://registry.npmjs.org/es5-ext/-/es5-ext-0.10.64.tgz",
+      "integrity": "sha512-p2snDhiLaXe6dahss1LddxqEm+SkuDvV8dnIQG0MWjyHpcMNfXKPE+/Cc0y+PhxJX3A4xGNeFCj5oc0BUh6deg==",
      "requires": {
        "es6-iterator": "^2.0.3",
        "es6-symbol": "^3.1.3",
+        "esniff": "^2.0.1",
        "next-tick": "^1.1.0"
      }
    },
@@ -9766,6 +9780,24 @@
      "integrity": "sha512-wpc+LXeiyiisxPlEkUzU6svyS1frIO3Mgxj1fdy7Pm8Ygzguax2N3Fa/D/ag1WqbOprdI+uY6wMUl8/a2G+iag==",
      "dev": true
    },
+    "esniff": {
+      "version": "2.0.1",
+      "resolved": "https://registry.npmjs.org/esniff/-/esniff-2.0.1.tgz",
+      "integrity": "sha512-kTUIGKQ/mDPFoJ0oVfcmyJn4iBDRptjNVIzwIFR7tqWXdVI9xfA2RMwY/gbSpJG3lkdWNEjLap/NqVHZiJsdfg==",
+      "requires": {
+        "d": "^1.0.1",
+        "es5-ext": "^0.10.62",
+        "event-emitter": "^0.3.5",
+        "type": "^2.7.2"
+      },
+      "dependencies": {
+        "type": {
+          "version": "2.7.2",
+          "resolved": "https://registry.npmjs.org/type/-/type-2.7.2.tgz",
+          "integrity": "sha512-dzlvlNlt6AXU7EBSfpAscydQ7gXB+pPGsPnfJnZpiNJBDj7IaJzQlBZYGdEi4R9HmPdBv2XmWJ6YUtoTa7lmCw=="
+        }
+      }
+    },
    "espree": {
      "version": "9.6.1",
      "resolved": "https://registry.npmjs.org/espree/-/espree-9.6.1.tgz",
@@ -9813,6 +9845,15 @@
      "integrity": "sha512-kVscqXk4OCp68SZ0dkgEKVi6/8ij300KBWTJq32P/dYeWTSwK41WyTxalN1eRmA5Z9UU/LX9D7FWSmV9SAYx6g==",
      "dev": true
    },
+    "event-emitter": {
+      "version": "0.3.5",
+      "resolved": "https://registry.npmjs.org/event-emitter/-/event-emitter-0.3.5.tgz",
+      "integrity": "sha512-D9rRn9y7kLPnJ+hMq7S/nhvoKwwvVJahBi2BPmx3bvbsEdK3W9ii8cBSGjP+72/LnM4n6fo3+dkCX5FeTQruXA==",
+      "requires": {
+        "d": "1",
+        "es5-ext": "~0.10.14"
+      }
+    },
    "event-target-shim": {
      "version": "5.0.1",
      "resolved": "https://registry.npmjs.org/event-target-shim/-/event-target-shim-5.0.1.tgz",
@@ -9964,9 +10005,9 @@
      "dev": true
    },
    "follow-redirects": {
-      "version": "1.15.5",
-      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.5.tgz",
-      "integrity": "sha512-vSFWUON1B+yAw1VN4xMfxgn5fTUiaOzAJCKBwIIgT/+7CuGy9+r+5gITvP62j3RmaD5Ph65UaERdOSRGUzZtgw=="
+      "version": "1.15.6",
+      "resolved": "https://registry.npmjs.org/follow-redirects/-/follow-redirects-1.15.6.tgz",
+      "integrity": "sha512-wWN62YITEaOpSK584EZXJafH1AGpO8RVgElfkuXbTOrPX4fIfOyEpW/CsiNd8JdYrAoOvafRTOEnvsO++qCqFA=="
    },
    "for-each": {
      "version": "0.3.3",
@@ -11108,9 +11149,9 @@
      }
    },
    "microsoft-cognitiveservices-speech-sdk": {
-      "version": "1.34.0",
-      "resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.34.0.tgz",
-      "integrity": "sha512-WAR0YqouRzVux2kI+f5wTPC6NyJgIVC1g65d79dJ9I32WPJs2kK+eb/BMB6mhSdCjackO5FsrW7JLaQ/vB1heQ==",
+      "version": "1.36.0",
+      "resolved": "https://registry.npmjs.org/microsoft-cognitiveservices-speech-sdk/-/microsoft-cognitiveservices-speech-sdk-1.36.0.tgz",
+      "integrity": "sha512-wPxuEXgjLdqMMIrdBtl8jquGahLV19LQE0ie8MI/PcBcNLG5buVzwS2rQEyHMsRGx+C/4OdBo1ROdNIUzCm4Lg==",
      "requires": {
        "@types/webrtc": "^0.0.37",
        "agent-base": "^6.0.1",
@@ -12405,12 +12446,9 @@
      }
    },
    "undici": {
-      "version": "6.4.0",
-      "resolved": "https://registry.npmjs.org/undici/-/undici-6.4.0.tgz",
-      "integrity": "sha512-wYaKgftNqf6Je7JQ51YzkEkEevzOgM7at5JytKO7BjaURQpERW8edQSMrr2xb+Yv4U8Yg47J24+lc9+NbeXMFA==",
-      "requires": {
-        "@fastify/busboy": "^2.0.0"
-      }
+      "version": "6.11.1",
+      "resolved": "https://registry.npmjs.org/undici/-/undici-6.11.1.tgz",
+      "integrity": "sha512-KyhzaLJnV1qa3BSHdj4AZ2ndqI0QWPxYzaIOio0WzcEJB9gvuysprJSLtpvc2D9mhR9jPDUk7xlJlZbH2KR5iw=="
    },
    "undici-types": {
      "version": "5.26.5",
@@ -1,6 +1,6 @@
 {
  "name": "@jambonz/speech-utils",
-  "version": "0.0.43",
+  "version": "0.0.49",
  "description": "TTS-related speech utilities for jambonz",
  "main": "index.js",
  "author": "Dave Horton",
@@ -34,7 +34,7 @@
    "form-urlencoded": "^6.1.4",
    "google-protobuf": "^3.21.2",
    "ibm-watson": "^8.0.0",
-    "microsoft-cognitiveservices-speech-sdk": "1.34.0",
+    "microsoft-cognitiveservices-speech-sdk": "1.36.0",
    "openai": "^4.25.0",
    "undici": "^6.4.0"
  },
@@ -188,6 +188,7 @@ test('Azure speech synth tests', async(t) => {
      language: 'en-US',
      voice: 'en-US-ChristopherNeural',
      text: longText,
+      renderForCaching: true
    });
    t.ok(!opts.servedFromCache, `successfully synthesized microsoft audio to ${opts.filePath}`);
    if (process.env.JAMBONES_HTTP_PROXY_IP && process.env.JAMBONES_HTTP_PROXY_PORT) {
@@ -203,6 +204,7 @@ test('Azure speech synth tests', async(t) => {
      language: 'en-US',
      voice: 'en-US-ChristopherNeural',
      text: longText,
+      renderForCaching: true
    });
    t.ok(opts.servedFromCache, `successfully retrieved microsoft audio from cache ${opts.filePath}`);
  } catch (err) {
@@ -212,6 +214,58 @@ test('Azure speech synth tests', async(t) => {
  client.quit();
 });

+test('Azure SSML tests', async(t) => {
+  const fn = require('..');
+  const {synthAudio, client} = fn(opts, logger);
+
+  if (!process.env.MICROSOFT_API_KEY || !process.env.MICROSOFT_REGION) {
+    t.pass('skipping Microsoft speech synth tests since MICROSOFT_API_KEY or MICROSOFT_REGION not provided');
+    return t.end();
+  }
+  try {
+    const text = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="en-US">
+    <voice name="en-US-JennyMultilingualNeural">
+    <mstts:express-as style="cheerful" styledegree="2">That'd be just amazing!
+    </mstts:express-as>
+    </voice>
+    </speak>`;
+
+    let opts = await synthAudio(stats, {
+      vendor: 'microsoft',
+      credentials: {
+        api_key: process.env.MICROSOFT_API_KEY,
+        region: process.env.MICROSOFT_REGION,
+      },
+      language: 'en-US',
+      voice: 'en-US-ChristopherNeural',
+      text,
+      renderForCaching: true
+    });
+    t.ok(!opts.servedFromCache, `successfully synthesized microsoft audio to ${opts.filePath}`);
+    if (process.env.JAMBONES_HTTP_PROXY_IP && process.env.JAMBONES_HTTP_PROXY_PORT) {
+      t.pass('successfully used proxy to reach microsoft tts service');
+    }
+
+    opts = await synthAudio(stats, {
+      vendor: 'microsoft',
+      credentials: {
+        api_key: process.env.MICROSOFT_API_KEY,
+        region: process.env.MICROSOFT_REGION,
+      },
+      language: 'en-US',
+      voice: 'en-US-ChristopherNeural',
+      text,
+      renderForCaching: true
+    });
+    t.ok(opts.servedFromCache, `successfully retrieved microsoft audio from cache ${opts.filePath}`);
+  } catch (err) {
+    console.error(err);
+    t.end(err);
+  }
+  client.quit();
+});
+
+
 test('Azure custom voice speech synth tests', async(t) => {
  const fn = require('..');
  const {synthAudio, client} = fn(opts, logger);
@@ -233,6 +287,7 @@ test('Azure custom voice speech synth tests', async(t) => {
      language: 'en-US',
      voice: process.env.MICROSOFT_CUSTOM_VOICE,
      text,
+      renderForCaching: true
    });
    t.ok(!opts.servedFromCache, `successfully synthesized microsoft audio to ${opts.filePath}`);

@@ -247,6 +302,7 @@ test('Azure custom voice speech synth tests', async(t) => {
      language: 'en-US',
      voice: process.env.MICROSOFT_CUSTOM_VOICE,
      text,
+      renderForCaching: true
    });
    t.ok(opts.servedFromCache, `successfully retrieved microsoft custom voice audio from cache ${opts.filePath}`);
  } catch (err) {
@@ -539,6 +595,7 @@ test('Deepgram speech synth tests', async(t) => {
      },
      model: 'aura-asteria-en',
      text,
+      renderForCaching: true
    });
    t.ok(!opts.servedFromCache, `successfully synthesized deepgram audio to ${opts.filePath}`);
Author	SHA1	Message	Date
Dave Horton	8650328e64	0.0.49	2024-04-07 12:15:02 -04:00
Dave Horton	85fa6da5b5	update to azure speech sdk 1.36.0	2024-04-07 12:14:55 -04:00
Dave Horton	e54e913fdd	0.0.48	2024-04-04 15:52:40 -04:00
Dave Horton	b088c0d7d9	Merge pull request #63 from jambonz/feat/mod_deepgram_tts Feat/mod deepgram tts	2024-04-04 15:52:10 -04:00
Hoan Luu Huu	f154a40692	Merge branch 'main' into feat/mod_deepgram_tts	2024-04-04 18:58:38 +07:00
Quan HL	0471b94ebe	wip	2024-04-04 10:12:24 +07:00
Dave Horton	8279891dff	0.0.47	2024-04-03 13:33:24 -04:00
Dave Horton	f546ca998d	cache files for azure tts streaming are r8	2024-04-03 13:32:32 -04:00
Dave Horton	bf229d0ab0	0.0.46	2024-04-03 13:22:20 -04:00
Dave Horton	f08fedb8ca	enable caching from azure tts streaming	2024-04-03 13:17:36 -04:00
Quan HL	f3cc38089c	mod_deepgra_tts	2024-04-03 20:46:52 +07:00
Dave Horton	fbed59e5de	0.0.45	2024-04-02 15:10:54 -04:00
Dave Horton	4f1685a365	Merge pull request #59 from jambonz/feat/azure_tts support azure streaming	2024-03-30 09:21:07 -04:00
Quan HL	2701af102a	wip	2024-03-30 17:49:14 +07:00
Quan HL	7f939b96d2	wip	2024-03-30 17:38:00 +07:00
Quan HL	4d58ca6daf	wip	2024-03-30 17:34:49 +07:00
Hoan Luu Huu	16dd7a2805	Merge branch 'main' into feat/azure_tts	2024-03-30 17:04:01 +07:00
Dave Horton	8f3e930004	0.0.44	2024-03-20 19:43:26 -04:00
Dave Horton	3f4c444d82	add azure SSML tests	2024-03-20 19:43:18 -04:00
Hoan Luu Huu	fd7d8b8bcd	Merge pull request #62 from jambonz/feat/mod_dub say command for freeswitch module to include vendor and voice	2024-03-20 13:40:51 +07:00
Hoan Luu Huu	46f833c7fa	Merge branch 'main' into feat/mod_dub	2024-03-12 18:10:21 +07:00
Quan HL	f3ab2baa6a	wip	2024-03-10 07:34:55 +07:00
Hoan Luu Huu	fb412e2ddf	Merge branch 'main' into feat/mod_dub	2024-03-10 06:43:09 +07:00
Quan HL	f06f96a6f0	wip	2024-03-10 06:41:46 +07:00
Hoan Luu Huu	2988e800b1	Merge branch 'main' into feat/azure_tts	2024-03-10 06:39:00 +07:00
Quan HL	c3188e40bb	support mod_dub	2024-03-09 16:59:00 +07:00
Quan HL	31a54f595b	wip	2024-02-26 15:42:09 +07:00
Quan HL	3560a6d4d9	wip	2024-02-26 14:01:53 +07:00
Quan HL	be8053db4f	wip	2024-02-26 13:54:34 +07:00
Quan HL	4ffae38a3f	wip	2024-02-26 13:49:37 +07:00
Quan HL	9e74760c39	support azure streaming	2024-02-26 13:33:22 +07:00