@andresaya/edge-tts 1.7.1 → 1.8.0

package/dist/index.js

@@ -1,5 +1,7 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.EdgeTTS = void 0;
+exports.Constants = exports.EdgeTTS = void 0;
 var EdgeTTS_1 = require("./services/EdgeTTS");
 Object.defineProperty(exports, "EdgeTTS", { enumerable: true, get: function () { return EdgeTTS_1.EdgeTTS; } });
+var constants_1 = require("./config/constants");
+Object.defineProperty(exports, "Constants", { enumerable: true, get: function () { return constants_1.Constants; } });

package/dist/services/EdgeTTS.d.ts

@@ -11,7 +11,6 @@ export interface SynthesisOptions {
     pitch?: string | number;
     rate?: string | number;
     volume?: string | number;
-    inputType?: 'auto' | 'ssml' | 'text';
     outputFormat?: string;
 }
 export interface WordBoundary {
@@ -26,6 +25,7 @@ export declare class EdgeTTS {
     private output_format;
     private word_boundaries;
     private ws;
+    normalizeVoices(data: any[]): Promise<Voice[]>;
     getVoices(): Promise<Voice[]>;
     getVoicesByLanguage(locale: string): Promise<Voice[]>;
     getVoicesByGender(gender: 'Male' | 'Female'): Promise<Voice[]>;
@@ -34,7 +34,6 @@ export declare class EdgeTTS {
     private validateRate;
     private validateVolume;
     synthesize(text: string, voice?: string, options?: SynthesisOptions): Promise<void>;
-    private detectSSML;
     private escapeXML;
     private getSSML;
     private nowRFC1123;

package/dist/services/EdgeTTS.js

@@ -30,30 +30,88 @@ class EdgeTTS {
     output_format = 'audio-24khz-48kbitrate-mono-mp3';
     word_boundaries = [];
     ws;
+    async normalizeVoices(data) {
+        const out = [];
+        for (const v of data || []) {
+            const short = v?.ShortName || "";
+            const locale = v?.Locale || "";
+            // base: remove locale prefix and Neural/NeuralHD suffix
+            let base = short.replace(/^[a-z]{2}-[A-Z]{2}-/, "");
+            base = base.replace(/NeuralHD$/, "").replace(/Neural$/, "").trim();
+            // VoiceType: if NeuralHD/Neural in Name or ShortName
+            const mix = `${v?.Name || ""} ${short}`;
+            const voiceType = v?.VoiceType || (/NeuralHD/i.test(mix) ? "NeuralHD" : "Neural");
+            // LocaleName: prefer LocaleName -> LanguageName -> locale
+            const localeName = v?.LocaleName || (locale || null);
+            // DisplayName: prefer DisplayName -> FriendlyName -> base -> short
+            let display = v?.DisplayName || v?.FriendlyName || base || short;
+            display = display.replace(/^Microsoft\s+/i, "");
+            display = display.split(" - ")[0].trim();
+            display = display.replace(/\s*Online\s*\(Natural\)\s*/i, " ");
+            display = display.replace(/\s*Online\s*/i, " ");
+            display = display.replace(/\s+/g, " ").trim();
+            // VoiceTag parsing
+            const tag = (v?.VoiceTag && typeof v.VoiceTag === "object") ? v.VoiceTag : {};
+            const tailored = Array.isArray(tag.TailoredScenarios)
+                ? tag.TailoredScenarios
+                : (Array.isArray(tag.ContentCategories) ? tag.ContentCategories : []);
+            const personalities = Array.isArray(tag.VoicePersonalities)
+                ? tag.VoicePersonalities
+                : [];
+            out.push({
+                Name: short || (v?.Name || ""),
+                DisplayName: display,
+                LocalName: display,
+                ShortName: short || (v?.Name || ""),
+                Gender: v?.Gender ?? null,
+                Locale: locale || null,
+                LocaleName: localeName,
+                SecondaryLocaleList: Array.isArray(v?.SecondaryLocaleList) ? v.SecondaryLocaleList : [],
+                VoiceType: voiceType,
+                VoiceTag: {
+                    TailoredScenarios: tailored,
+                    VoicePersonalities: personalities,
+                },
+                FriendlyName: `${display} (${voiceType}) - ${localeName}`,
+            });
+        }
+        return out;
+    }
     async getVoices() {
         const secMsGEC = await this.generateSecMsGec(constants_1.Constants.TRUSTED_CLIENT_TOKEN);
-        const httpsAgent = new https_1.default.Agent({
-            rejectUnauthorized: false,
-        });
-        const response = await new Promise((resolve, reject) => {
-            const req = https_1.default.request(`${constants_1.Constants.VOICES_URL}?Ocp-Apim-Subscription-Key=${constants_1.Constants.TRUSTED_CLIENT_TOKEN}&Sec-MS-GEC=${secMsGEC}&Sec-MS-GEC-Version=${constants_1.Constants.VERSION_MS_GEC}`, {
-                headers: constants_1.Constants.getBaseHeaders(),
-                agent: httpsAgent
-            }, (res) => {
-                let data = '';
-                res.on('data', (chunk) => (data += chunk));
-                res.on('end', () => resolve(new Response(data, { status: res.statusCode || 200 })));
+        const httpsAgent = new https_1.default.Agent({ rejectUnauthorized: false });
+        const url = `${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(),
+            "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 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");
+                    try {
+                        const parsed = JSON.parse(body);
+                        // soporta array directo o { voices: [...] }
+                        const voices = Array.isArray(parsed)
+                            ? parsed
+                            : (parsed.voices || parsed.Voices || []);
+                        resolve(Array.isArray(voices) ? voices : []);
+                    }
+                    catch (e) {
+                        reject(new Error("JSON inválido: " + (e?.message || String(e))));
+                    }
+                });
             });
-            req.on('error', reject);
+            req.on("error", reject);
             req.end();
         });
-        const data = await response.json();
-        return data.map((voice) => {
-            voice.FriendlyName = voice.FriendlyName || voice.LocalName;
-            delete voice.SampleRateHertz;
-            delete voice.Status;
-            return voice;
-        });
+        return this.normalizeVoices(voicesRaw);
     }
     async getVoicesByLanguage(locale) {
         const voices = await this.getVoices();
@@ -114,7 +172,7 @@ class EdgeTTS {
         return new Promise((resolve, reject) => {
             this.audio_stream = [];
             const reqId = this.generateUUID();
-            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}`;
+            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}`;
             this.ws = new ws_1.default(url, {
                 headers: constants_1.Constants.getBaseHeaders(),
                 rejectUnauthorized: false
@@ -161,31 +219,6 @@ class EdgeTTS {
             });
         });
     }
-    detectSSML(content) {
-        const trimmedContent = content.trim();
-        const looksLikeSSML = /^<\?xml|^<speak/i.test(trimmedContent);
-        if (!looksLikeSSML) {
-            return { isValid: true, isSSML: false };
-        }
-        const errors = [];
-        const hasSpeakTag = /<speak\b[^>]*>[\s\S]*<\/speak>/i.test(trimmedContent);
-        const hasVoiceTag = /<voice\b[^>]*>[\s\S]*<\/voice>/i.test(trimmedContent);
-        if (!hasSpeakTag) {
-            throw new Error('Invalid SSML: Missing <speak> tag');
-        }
-        if (!hasVoiceTag) {
-            throw new Error('Invalid SSML: Missing <voice> tag');
-        }
-        const hasCorrectNamespace = /xmlns="http:\/\/www\.w3\.org\/2001\/10\/synthesis"/i.test(trimmedContent);
-        if (!hasCorrectNamespace && hasSpeakTag) {
-            throw new Error('Invalid SSML: Missing or incorrect namespace declaration');
-        }
-        return {
-            isValid: errors.length === 0,
-            isSSML: hasSpeakTag || hasVoiceTag,
-            errors: errors.length > 0 ? errors : undefined
-        };
-    }
     escapeXML(text) {
         return text
             .replace(/&/g, '&amp;')
@@ -195,37 +228,6 @@ class EdgeTTS {
             .replace(/'/g, '&apos;');
     }
     getSSML(content, voice, options = {}) {
-        const inputType = options.inputType || 'auto';
-        let treatAsSSML = false;
-        if (inputType === 'ssml') {
-            treatAsSSML = true;
-        }
-        else if (inputType === 'text') {
-            treatAsSSML = false;
-        }
-        else { // 'auto'
-            const detection = this.detectSSML(content);
-            treatAsSSML = detection.isSSML;
-            if (detection.isSSML) {
-                console.log('→ Detected SSML input');
-                if (!detection.isValid) {
-                    console.warn('⚠ SSML validation warnings:', detection.errors);
-                }
-            }
-            else {
-                console.log('→ Detected plain text input');
-            }
-        }
-        if (treatAsSSML) {
-            let ssml = content.trim();
-            if (!ssml.includes('xmlns=')) {
-                ssml = ssml.replace(/<speak([^>]*)>/i, '<speak$1 xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="https://www.w3.org/2001/mstts">');
-            }
-            if (!/<voice\b[^>]*>/i.test(ssml) && voice) {
-                ssml = ssml.replace(/(<speak[^>]*>)([\s\S]*?)(<\/speak>)/i, `$1<voice name="${voice}">$2</voice>$3`);
-            }
-            return ssml;
-        }
         const pitch = this.validatePitch(options.pitch ?? 0);
         const rate = this.validateRate(options.rate ?? 0);
         const volume = this.validateVolume(options.volume ?? 0);
@@ -266,7 +268,7 @@ class EdgeTTS {
         this.audio_stream = [];
         const reqId = this.generateUUID();
         const secMsGEC = await this.generateSecMsGec(constants_1.Constants.TRUSTED_CLIENT_TOKEN);
-        const url = `${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}`;
+        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}`;
         this.ws = new ws_1.default(url, {
             headers: constants_1.Constants.getBaseHeaders(),
             rejectUnauthorized: false

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@andresaya/edge-tts",
-  "version": "1.7.1",
+  "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",
@@ -46,7 +46,7 @@
     "dist"
   ],
   "bin": {
-    "edge-tts": "./dist/cli/edge-tts.js"
+    "edge-tts": "dist/cli/edge-tts.js"
   },
   "engines": {
     "bun": ">=0.5.0"