@andresaya/edge-tts 1.8.0 → 1.8.3

package/dist/services/EdgeTTS.js

@@ -79,22 +79,26 @@ class EdgeTTS {
     }
     async getVoices() {
         const secMsGEC = await this.generateSecMsGec(constants_1.Constants.TRUSTED_CLIENT_TOKEN);
-        const httpsAgent = new https_1.default.Agent({ rejectUnauthorized: false });
         const url = `${constants_1.Constants.VOICES_URL}` +
             `?TrustedClientToken=${constants_1.Constants.TRUSTED_CLIENT_TOKEN}` +
             `&Sec-MS-GEC=${secMsGEC}` +
             `&Sec-MS-GEC-Version=${constants_1.Constants.VERSION_MS_GEC}`;
         const headers = {
             ...constants_1.Constants.getBaseHeaders(),
+            ...constants_1.Constants.VOICE_HEADERS,
             "Accept-Encoding": "identity", // evita gzip/br/zstd
         };
         const voicesRaw = await new Promise((resolve, reject) => {
-            const req = https_1.default.request(url, { method: "GET", headers, agent: httpsAgent }, (res) => {
+            const req = https_1.default.request(url, { method: "GET", headers }, (res) => {
                 const chunks = [];
                 res.on("data", (chunk) => chunks.push(buffer_1.Buffer.isBuffer(chunk) ? chunk : buffer_1.Buffer.from(chunk)));
                 res.on("error", reject);
                 res.on("end", () => {
                     const body = buffer_1.Buffer.concat(chunks).toString("utf8");
+                    if (res.statusCode !== 200) {
+                        reject(new Error(`Voice list request failed with status ${res.statusCode}`));
+                        return;
+                    }
                     try {
                         const parsed = JSON.parse(body);
                         // soporta array directo o { voices: [...] }
@@ -104,7 +108,7 @@ class EdgeTTS {
                         resolve(Array.isArray(voices) ? voices : []);
                     }
                     catch (e) {
-                        reject(new Error("JSON inválido: " + (e?.message || String(e))));
+                        reject(new Error("Invalid voice list JSON: " + (e?.message || String(e))));
                     }
                 });
             });
@@ -163,7 +167,7 @@ class EdgeTTS {
             volumeValue = volume;
         }
         if (volumeValue < -100 || volumeValue > 100) {
-            throw new Error("Volume cannot be negative. Expected a value from -100% to 100% (or more).");
+            throw new Error("Invalid volume. Expected a value from -100% to 100%.");
         }
         return `${volumeValue}%`;
     }
@@ -171,11 +175,11 @@ class EdgeTTS {
         const secMsGEC = await this.generateSecMsGec(constants_1.Constants.TRUSTED_CLIENT_TOKEN);
         return new Promise((resolve, reject) => {
             this.audio_stream = [];
+            this.word_boundaries = [];
             const reqId = this.generateUUID();
-            const url = `${constants_1.Constants.WSS_URL}?TrustedClientToken=${constants_1.Constants.TRUSTED_CLIENT_TOKEN}&Sec-MS-GEC=${secMsGEC}&Sec-MS-GEC-Version=${constants_1.Constants.VERSION_MS_GEC}&ConnectionId=${reqId}`;
+            const url = `${constants_1.Constants.WSS_URL}?Ocp-Apim-Subscription-Key=${constants_1.Constants.TRUSTED_CLIENT_TOKEN}&Sec-MS-GEC=${secMsGEC}&Sec-MS-GEC-Version=${constants_1.Constants.VERSION_MS_GEC}&ConnectionId=${reqId}`;
             this.ws = new ws_1.default(url, {
-                headers: constants_1.Constants.getBaseHeaders(),
-                rejectUnauthorized: false
+                headers: { ...constants_1.Constants.getBaseHeaders(), ...constants_1.Constants.WSS_HEADERS }
             });
             const SSML_text = this.getSSML(text, voice, options);
             const outputFormat = options.outputFormat || 'audio-24khz-48kbitrate-mono-mp3';
@@ -232,13 +236,13 @@ class EdgeTTS {
         const rate = this.validateRate(options.rate ?? 0);
         const volume = this.validateVolume(options.volume ?? 0);
         const escapedText = this.escapeXML(content);
-        return `<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="${voice}">
-                        <prosody pitch="${pitch}" rate="${rate}" volume="${volume}">
-                            ${escapedText}
-                        </prosody>
-                    </voice>
-                </speak>
+        return `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts" xml:lang="en-US">
+                    <voice name="${voice}">
+                        <prosody pitch="${pitch}" rate="${rate}" volume="${volume}">
+                            ${escapedText}
+                        </prosody>
+                    </voice>
+                </speak>
         `;
     }
     nowRFC1123(timeZone = 'UTC') {
@@ -266,12 +270,12 @@ class EdgeTTS {
     }
     async *synthesizeStream(text, voice = 'en-US-AnaNeural', options = {}) {
         this.audio_stream = [];
+        this.word_boundaries = [];
         const reqId = this.generateUUID();
         const secMsGEC = await this.generateSecMsGec(constants_1.Constants.TRUSTED_CLIENT_TOKEN);
-        const url = `${constants_1.Constants.WSS_URL}?TrustedClientToken=${constants_1.Constants.TRUSTED_CLIENT_TOKEN}&Sec-MS-GEC=${secMsGEC}&Sec-MS-GEC-Version=${constants_1.Constants.VERSION_MS_GEC}&ConnectionId=${reqId}`;
+        const url = `${constants_1.Constants.WSS_URL}?Ocp-Apim-Subscription-Key=${constants_1.Constants.TRUSTED_CLIENT_TOKEN}&Sec-MS-GEC=${secMsGEC}&Sec-MS-GEC-Version=${constants_1.Constants.VERSION_MS_GEC}&ConnectionId=${reqId}`;
         this.ws = new ws_1.default(url, {
-            headers: constants_1.Constants.getBaseHeaders(),
-            rejectUnauthorized: false
+            headers: { ...constants_1.Constants.getBaseHeaders(), ...constants_1.Constants.WSS_HEADERS }
         });
         const SSML_text = this.getSSML(text, voice, options);
         const outputFormat = options.outputFormat || 'audio-24khz-48kbitrate-mono-mp3';
@@ -431,10 +435,10 @@ class EdgeTTS {
         if (this.audio_stream.length === 0) {
             throw new Error("No audio data available");
         }
-        // Estimate duration based on the size of the audio stream
+        // NOTE: MP3 is CBR; formats without a bitrate (opus) assume 48 kbit/s.
         const bufferSize = this.toBuffer().length;
-        const estimatedDuration = bufferSize / (24000 * 3); // 24000 Hz sample rate, 3 bytes per sample (16-bit stereo)
-        return estimatedDuration;
+        const kbps = Number(/(\d+)kbitrate/.exec(this.output_format)?.[1] ?? 48);
+        return (bufferSize * 8) / (kbps * 1000);
     }
     getFileExtension(format) {
         if (format.includes('mp3'))
@@ -486,7 +490,16 @@ class EdgeTTS {
         if (this.audio_stream.length === 0) {
             throw new Error("No audio data available. Did you run synthesize() first?");
         }
-        return buffer_1.Buffer.concat(this.audio_stream);
+        // When the service cannot speak the text it answers with a single empty
+        // chunk, so the stream is not empty but the audio is. Checking the byte
+        // count rather than the chunk count is what catches that.
+        const buffer = buffer_1.Buffer.concat(this.audio_stream);
+        if (buffer.length === 0) {
+            throw new Error("The service returned no audio. Voices are locale specific, so a voice cannot " +
+                "read text written in another script. Check that the voice matches the language " +
+                "of the text.");
+        }
+        return buffer;
     }
     async saveMetadata(outputPath) {
         if (this.word_boundaries.length === 0) {

package/package.json

@@ -1,54 +1,48 @@
-{
-  "name": "@andresaya/edge-tts",
-  "version": "1.8.0",
-  "description": "Edge TTS is a package that allows access to the online text-to-speech service used by Microsoft Edge without the need for Microsoft Edge, Windows, or an API key.",
-  "main": "dist/index.js",
-  "browser": "dist/browser/edge-tts.esm.js",
-  "unpkg": "dist/browser/edge-tts.umd.min.js",
-  "jsdelivr": "dist/browser/edge-tts.umd.min.js",
-  "keywords": [
-    "edge",
-    "tts",
-    "text-to-speech",
-    "microsoft",
-    "microsoft edge",
-    "speech-synthesis"
-  ],
-  "author": "Andres Aya",
-  "license": "GPL-3.0-only",
-  "homepage": "https://github.com/andresayac/edge-tts",
-  "bugs": {
-    "url": "https://github.com/andresayac/edge-tts/issues"
-  },
-  "scripts": {
-    "build": "tsc",
-    "build:bin": "bun scripts/build.ts",
-    "build:browser": "bun scripts/build-browser.ts",
-    "prepublishOnly": "npm run build && npm run build:bin && npm run build:browser"
-  },
-  "devDependencies": {
-    "@types/bun": "latest",
-    "@types/ws": "^8.18.1",
-    "esbuild": "^0.25.12",
-    "typescript": "^5.6.2"
-  },
-  "dependencies": {
-    "@types/uuid": "^10.0.0",
-    "commander": "^12.1.0",
-    "events": "^3.3.0",
-    "fs": "^0.0.1-security",
-    "ws": "^8.18.3"
-  },
-  "peerDependencies": {
-    "typescript": "^5.0.0"
-  },
-  "files": [
-    "dist"
-  ],
-  "bin": {
-    "edge-tts": "dist/cli/edge-tts.js"
-  },
-  "engines": {
-    "bun": ">=0.5.0"
-  }
-}
+{
+  "name": "@andresaya/edge-tts",
+  "version": "1.8.3",
+  "description": "Edge TTS is a package that allows access to the online text-to-speech service used by Microsoft Edge without the need for Microsoft Edge, Windows, or an API key.",
+  "main": "dist/index.js",
+  "types": "dist/index.d.ts",
+  "browser": "dist/browser/edge-tts.esm.js",
+  "unpkg": "dist/browser/edge-tts.umd.min.js",
+  "jsdelivr": "dist/browser/edge-tts.umd.min.js",
+  "keywords": [
+    "edge",
+    "tts",
+    "text-to-speech",
+    "microsoft",
+    "microsoft edge",
+    "speech-synthesis"
+  ],
+  "author": "Andres Aya",
+  "license": "GPL-3.0-only",
+  "homepage": "https://github.com/andresayac/edge-tts",
+  "bugs": {
+    "url": "https://github.com/andresayac/edge-tts/issues"
+  },
+  "devDependencies": {
+    "@types/bun": "latest",
+    "@types/ws": "^8.18.1",
+    "esbuild": "^0.25.12",
+    "typescript": "^5.6.2"
+  },
+  "dependencies": {
+    "commander": "^12.1.0",
+    "ws": "^8.18.3"
+  },
+  "files": [
+    "dist"
+  ],
+  "bin": {
+    "edge-tts": "dist/cli/edge-tts.js"
+  },
+  "engines": {
+    "bun": ">=0.5.0"
+  },
+  "scripts": {
+    "build": "tsc",
+    "build:bin": "bun scripts/build.ts",
+    "build:browser": "bun scripts/build-browser.ts"
+  }
+}