add support for azure custom voices on a per-say basis (#346)

lib/session/call-session.js

@@ -1291,6 +1291,7 @@ class CallSession extends Emitter {
       this.wakeupResolver(resolution);
       this.wakeupResolver = null;
     }
+    /*
     else {
       const {queue, command} = resolution;
       const {span} = this.rootSpan.startChildSpan(`recv cmd: ${command}`);
@@ -1300,6 +1301,7 @@ class CallSession extends Emitter {
       });
       span.end();
     }
+    */
   }
 
   _onWsConnectionDropped() {

lib/tasks/say.js

@@ -36,6 +36,7 @@ class TaskSay extends Task {
     this.earlyMedia = this.data.earlyMedia === true || (parentTask && parentTask.earlyMedia);
     this.synthesizer = this.data.synthesizer || {};
     this.disableTtsCache = this.data.disableTtsCache;
+    this.options = this.synthesizer.options || {};
   }
 
   get name() { return TaskName.Say; }
@@ -66,7 +67,7 @@ class TaskSay extends Task {
       cs.speechSynthesisVoice;
     const engine = this.synthesizer.engine || 'standard';
     const salt = cs.callSid;
-    const credentials = cs.getSpeechCredentials(vendor, 'tts');
+    let credentials = cs.getSpeechCredentials(vendor, 'tts');
 
     /* parse Nuance voices into name and model */
     let model;
@@ -78,6 +79,16 @@ class TaskSay extends Task {
       }
     }
 
+    /* allow for microsoft custom region voice and api_key to be specified as an override */
+    if (vendor === 'microsoft' && this.options.deploymentId) {
+      credentials = credentials || {};
+      credentials.use_custom_tts = true;
+      credentials.custom_tts_endpoint = this.options.deploymentId;
+      credentials.api_key = this.options.apiKey || credentials.apiKey;
+      credentials.region = this.options.region || credentials.region;
+      voice = this.options.voice || voice;
+    }
+
     this.logger.info({vendor, language, voice, model}, 'TaskSay:exec');
     this.ep = ep;
     try {

package-lock.json

@@ -15,7 +15,7 @@
         "@jambonz/speech-utils": "^0.0.12",
         "@jambonz/stats-collector": "^0.1.8",
         "@jambonz/time-series": "^0.2.5",
-        "@jambonz/verb-specifications": "^0.0.21",
+        "@jambonz/verb-specifications": "^0.0.22",
         "@opentelemetry/api": "^1.4.0",
         "@opentelemetry/exporter-jaeger": "^1.9.0",
         "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",
@@ -1676,9 +1676,9 @@
       }
     },
     "node_modules/@jambonz/verb-specifications": {
-      "version": "0.0.21",
+      "version": "0.0.22",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.21.tgz",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.22.tgz",
-      "integrity": "sha512-NQTQLI0k7nFmjqiKIbPaux2+ES2etFh/hnMNzPbszXS313mAnMZtv7y+EWrbniXeiXMCwHbkavm1H/N8+Lv5Qg==",
+      "integrity": "sha512-dgvpVh48IX2CVedLKgxYhv5ZwWoeAPtEhCjXqkkVOGgh6eqX1eu+pmcH2D/aKeFdxhXRVGUxfhtLLerfErIpVQ==",
       "dependencies": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"
@@ -10128,9 +10128,9 @@
       }
     },
     "@jambonz/verb-specifications": {
-      "version": "0.0.21",
+      "version": "0.0.22",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.21.tgz",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.22.tgz",
-      "integrity": "sha512-NQTQLI0k7nFmjqiKIbPaux2+ES2etFh/hnMNzPbszXS313mAnMZtv7y+EWrbniXeiXMCwHbkavm1H/N8+Lv5Qg==",
+      "integrity": "sha512-dgvpVh48IX2CVedLKgxYhv5ZwWoeAPtEhCjXqkkVOGgh6eqX1eu+pmcH2D/aKeFdxhXRVGUxfhtLLerfErIpVQ==",
       "requires": {
         "debug": "^4.3.4",
         "pino": "^8.8.0"

package.json

@@ -31,7 +31,7 @@
     "@jambonz/speech-utils": "^0.0.12",
     "@jambonz/stats-collector": "^0.1.8",
     "@jambonz/time-series": "^0.2.5",
-    "@jambonz/verb-specifications": "^0.0.21",
+    "@jambonz/verb-specifications": "^0.0.22",
     "@opentelemetry/api": "^1.4.0",
     "@opentelemetry/exporter-jaeger": "^1.9.0",
     "@opentelemetry/exporter-trace-otlp-http": "^0.35.0",

test/gather-tests.js

@@ -261,7 +261,7 @@ test('\'gather\' test - deepgram', async(t) => {
     await sippUac('uac-gather-account-creds-success.xml', '172.38.0.10', from);
     let obj = await getJSON(`http://127.0.0.1:3100/lastRequest/${from}_actionHook`);
     //console.log(JSON.stringify(obj));
-    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().startsWith('i\'d like to speak to customer support'),
+    t.ok(obj.body.speech.alternatives[0].transcript.toLowerCase().includes('like to speak to customer support'),
       'gather: succeeds when using deepgram credentials');
     disconnect();
   } catch (err) {

test/say-tests.js

@@ -83,3 +83,44 @@ test('\'config\' reset synthesizer tests', async(t) => {
     t.error(err);
   }
 });
+
+const {MICROSOFT_CUSTOM_API_KEY, MICROSOFT_DEPLOYMENT_ID, MICROSOFT_CUSTOM_REGION, MICROSOFT_CUSTOM_VOICE} = process.env;
+if (MICROSOFT_CUSTOM_API_KEY && MICROSOFT_DEPLOYMENT_ID && MICROSOFT_CUSTOM_REGION && MICROSOFT_CUSTOM_VOICE) {
+  test('\'say\' tests - microsoft custom voice', async(t) => {
+    clearModule.all();
+    const {srf, disconnect} = require('../app');
+
+    try {
+      await connect(srf);
+
+      // GIVEN
+      const verbs = [
+        {
+          verb: 'say',
+          text: 'hello',
+          synthesizer: {
+            vendor: 'microsoft',
+            voice: MICROSOFT_CUSTOM_VOICE,
+            options: {
+              deploymentId: MICROSOFT_DEPLOYMENT_ID,
+              apiKey: MICROSOFT_CUSTOM_API_KEY,
+              region: MICROSOFT_CUSTOM_REGION,
+            }
+          }
+        }
+      ];
+
+      const from = 'say_test_success';
+      provisionCallHook(from, verbs)
+
+      // THEN
+      await sippUac('uac-success-received-bye.xml', '172.38.0.10', from);
+      t.pass('say: succeeds when using microsoft custom voice');
+      disconnect();
+    } catch (err) {
+      console.log(`error received: ${err}`);
+      disconnect();
+      t.error(err);
+    }
+  });
+}