fix(azure): nest <prosody> inside <voice>, not as direct child of <speak> (#39)

When rate, pitch, or volume were passed as options to ensureAzureSSMLStructure,
the prosody element was being extracted from inside <speak> rather than from
inside <voice>. This produced invalid SSML that Azure rejects with:

  "Node [speak] with type [RootSpeak] should not contain node [prosody]
   with type [Others]."

Fix: when a <voice> element is present, replace content inside <voice>
(mirroring the existing this.properties branch above). Add regression test.

Fixes #38

Author: OwenMcGirr

src/__tests__/azure-mstts-namespace.test.ts

@@ -128,5 +128,26 @@ describe("Azure MSTTS Namespace Handling", () => {
       const xmlnsMatches = result.match(/xmlns="http:\/\/www\.w3\.org\/2001\/10\/synthesis"/g);
       expect(xmlnsMatches?.length).toBe(1);
     });
+
+    it("should nest <prosody> inside <voice>, not as a direct child of <speak>", async () => {
+      // Regression test for: https://github.com/willwade/js-tts-wrapper/issues/38
+      // When rate/pitch/volume are passed as options, <prosody> was placed outside
+      // <voice>, which Azure rejects with:
+      //   "Node [speak] with type [RootSpeak] should not contain node [prosody] with type [Others]"
+      const plainSSML = `<speak>Hello world</speak>`;
+      const options = { rate: "fast", pitch: "high", volume: 80 };
+
+      const result = (client as any).ensureAzureSSMLStructure(plainSSML, "en-US-JennyNeural", options);
+
+      // <prosody> must appear after <voice>, not before it
+      const voiceIndex = result.indexOf("<voice");
+      const prosodyIndex = result.indexOf("<prosody");
+      expect(voiceIndex).toBeGreaterThan(-1);
+      expect(prosodyIndex).toBeGreaterThan(voiceIndex);
+
+      // The structure must be <speak><voice><prosody>...</prosody></voice></speak>
+      expect(result).toMatch(/<voice[^>]*>\s*<prosody[^>]*>/);
+      expect(result).toMatch(/<\/prosody>\s*<\/voice>/);
+    });
   });
 });

src/engines/azure.ts

@@ -646,12 +646,22 @@ export class AzureTTSClient extends AbstractTTSClient {
       if (options.volume !== undefined) attrs.push(`volume="${options.volume}%"`);
 
       if (attrs.length > 0) {
-        // Extract content
-        const match = ssml.match(/<speak[^>]*>(.*?)<\/speak>/s);
-        if (match) {
-          const content = match[1];
-          const prosodyContent = `<prosody ${attrs.join(" ")}>${content}</prosody>`;
-          ssml = ssml.replace(content, prosodyContent);
+        // Extract content from inside <voice> if present, otherwise from <speak>.
+        // Prosody must be nested inside <voice>, not as a direct child of <speak>.
+        if (ssml.includes("<voice")) {
+          const match = ssml.match(/<voice[^>]*>(.*?)<\/voice>/s);
+          if (match) {
+            const content = match[1];
+            const prosodyContent = `<prosody ${attrs.join(" ")}>${content}</prosody>`;
+            ssml = ssml.replace(content, prosodyContent);
+          }
+        } else {
+          const match = ssml.match(/<speak[^>]*>(.*?)<\/speak>/s);
+          if (match) {
+            const content = match[1];
+            const prosodyContent = `<prosody ${attrs.join(" ")}>${content}</prosody>`;
+            ssml = ssml.replace(content, prosodyContent);
+          }
         }
       }
     }