@andresaya/edge-tts 1.6.2 → 1.7.1

package/dist/services/EdgeTTS.js

@@ -27,6 +27,7 @@ function ensureBuffer(data) {
 class EdgeTTS {
     audio_stream = [];
     audio_format = 'mp3';
+    output_format = 'audio-24khz-48kbitrate-mono-mp3';
     word_boundaries = [];
     ws;
     async getVoices() {
@@ -119,6 +120,8 @@ class EdgeTTS {
                 rejectUnauthorized: false
             });
             const SSML_text = this.getSSML(text, voice, options);
+            const outputFormat = options.outputFormat || 'audio-24khz-48kbitrate-mono-mp3';
+            this.output_format = outputFormat;
             let timedOut = false;
             let inactivityTimeout;
             const resetInactivityTimeout = () => {
@@ -133,7 +136,7 @@ class EdgeTTS {
             };
             this.ws.on('open', () => {
                 resetInactivityTimeout(); // start the inactivity timeout
-                const message = this.buildTTSConfigMessage();
+                const message = this.buildTTSConfigMessage(outputFormat);
                 this.ws.send(message);
                 const timestamp = this.nowRFC1123();
                 const speechMessage = `X-RequestId:${reqId}\r\nContent-Type:application/ssml+xml\r\nX-Timestamp:${timestamp}\r\nPath:ssml\r\n\r\n${SSML_text}`;
@@ -254,10 +257,10 @@ class EdgeTTS {
     parseRFC1123(rfcStr) {
         return new Date(rfcStr);
     }
-    buildTTSConfigMessage() {
+    buildTTSConfigMessage(outputFormat = 'audio-24khz-48kbitrate-mono-mp3') {
         const timestamp = this.nowRFC1123();
         return `X-Timestamp:${timestamp}\r\nContent-Type:application/json; charset=utf-8\r\nPath:speech.config\r\n\r\n` +
-            `{"context":{"synthesis":{"audio":{"metadataoptions":{"sentenceBoundaryEnabled":false,"wordBoundaryEnabled":true},"outputFormat":"audio-24khz-48kbitrate-mono-mp3"}}}}`;
+            `{"context":{"synthesis":{"audio":{"metadataoptions":{"sentenceBoundaryEnabled":false,"wordBoundaryEnabled":true},"outputFormat":"${outputFormat}"}}}}`;
     }
     async *synthesizeStream(text, voice = 'en-US-AnaNeural', options = {}) {
         this.audio_stream = [];
@@ -269,6 +272,8 @@ class EdgeTTS {
             rejectUnauthorized: false
         });
         const SSML_text = this.getSSML(text, voice, options);
+        const outputFormat = options.outputFormat || 'audio-24khz-48kbitrate-mono-mp3';
+        this.output_format = outputFormat;
         const queue = [];
         let done = false;
         let error = null;
@@ -299,7 +304,7 @@ class EdgeTTS {
         };
         this.ws.on('open', () => {
             resetInactivityTimeout(); // start the inactivity timeout
-            const message = this.buildTTSConfigMessage();
+            const message = this.buildTTSConfigMessage(outputFormat);
             this.ws.send(message);
             const timestamp = this.nowRFC1123();
             const speechMessage = `X-RequestId:${reqId}\r\nContent-Type:application/ssml+xml\r\nX-Timestamp:${timestamp}\r\nPath:ssml\r\n\r\n${SSML_text}`;
@@ -429,17 +434,41 @@ class EdgeTTS {
         const estimatedDuration = bufferSize / (24000 * 3); // 24000 Hz sample rate, 3 bytes per sample (16-bit stereo)
         return estimatedDuration;
     }
+    getFileExtension(format) {
+        if (format.includes('mp3'))
+            return 'mp3';
+        if (format.includes('opus') && format.includes('webm'))
+            return 'webm';
+        if (format.includes('opus') && format.includes('ogg'))
+            return 'ogg';
+        if (format.includes('wav') || format.includes('riff'))
+            return 'wav';
+        if (format.includes('pcm') && format.includes('raw'))
+            return 'pcm';
+        if (format.includes('alaw'))
+            return 'alaw';
+        if (format.includes('mulaw'))
+            return 'mulaw';
+        if (format.includes('truesilk'))
+            return 'silk';
+        if (format.includes('g722'))
+            return 'g722';
+        if (format.includes('amr'))
+            return 'amr';
+        return 'audio';
+    }
     getAudioInfo() {
         const buffer = this.toBuffer();
         return {
             size: buffer.length,
-            format: this.audio_format,
+            format: this.getFileExtension(this.output_format),
             estimatedDuration: this.getDuration()
         };
     }
-    async toFile(outputPath, format = this.audio_format) {
-        if (!format || typeof format !== 'string')
-            format = this.audio_format;
+    async toFile(outputPath, format) {
+        if (!format) {
+            format = this.getFileExtension(this.output_format);
+        }
         const audioBuffer = this.toBuffer();
         const finalPath = `${outputPath}.${format}`;
         await (0, promises_1.writeFile)(finalPath, new Uint8Array(audioBuffer));

package/package.json

@@ -1,54 +1,54 @@
-{
-  "name": "@andresaya/edge-tts",
-  "version": "1.6.2",
-  "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": "1.7.1",
+  "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"
+  }
+}

package/dist/browser.d.ts

@@ -1 +0,0 @@
-export { EdgeTTS, type Voice, type SynthesisOptions, type WordBoundary } from './browser/EdgeTTS.browser';

package/dist/browser.js

@@ -1,5 +0,0 @@
-"use strict";
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.EdgeTTS = void 0;
-var EdgeTTS_browser_1 = require("./browser/EdgeTTS.browser");
-Object.defineProperty(exports, "EdgeTTS", { enumerable: true, get: function () { return EdgeTTS_browser_1.EdgeTTS; } });