feat: add ElevenLabs Firebase token emulation

docker-compose.yml

@@ -8,7 +8,7 @@ services:
           DISCORD_TOKEN: 
           TTS_TIKTOK_SESSIONID: 
           TTS_ELEVENLABS_KEY: 
-          TTS_ELEVENLABS_TOKEN: 
+          TTS_ELEVENLABS_REFRESHTOKEN: 
       restart: unless-stopped
       volumes:
         - ./db.sqlite:/app/db.sqlite

src/modules/tts-modes/elevenlabs.ts

@@ -5,6 +5,10 @@ import * as https from 'https';
 
 const ELEVENLABS_API_ENDPOINT = 'api.elevenlabs.io';
 
+const FIREBASE_API_KEY = "AIzaSyBSsRE_1Os04-bxpd5JTLIniy3UK4OqKys";
+const FIREBASE_URL = `https://securetoken.googleapis.com/v1/token?key=${FIREBASE_API_KEY}`;
+
+
 /*
 	TO-DO: Implement previous text
 */
@@ -39,6 +43,12 @@ interface ElevenLabsStreamRequest {
 	voice_settings: ElevenLabsVoiceSettings;
 }
 
+interface FirebaseSession {
+	idToken: string;
+	refreshToken: string;
+	expiresAt: number;
+}
+
 export class ElevenLabsTTS implements TTSModule {
 	private voices: Array<ElevenLabsVoice> | undefined = undefined;
 	private models: Array<ElevenLabsModel> | undefined = undefined;
@@ -48,29 +58,38 @@ export class ElevenLabsTTS implements TTSModule {
 	public settings: ElevenLabsVoiceSettings;
 	public modelId: string;
 
+	private session: FirebaseSession | undefined = undefined;
+
+	private initializationPromise: Promise<void> | undefined = undefined;
+
 	public static readonly DEFAULT_SETTINGS: ElevenLabsVoiceSettings = {
-		stability: 0.5,
+		stability: 0.0,
 		similarity_boost: 0.5,
-		style: 0.0,
+		style: 1.0,
 		speed: 1.0,
 		user_speaker_boost: true
 	};
 
 	constructor() {
-		if (this.canBeUsed()) {
-			this.fetchVoices();
-			this.fetchModels();
-		}
-
 		this.settings = ElevenLabsTTS.DEFAULT_SETTINGS;
-
 		this.modelId = 'eleven_flash_v2_5';
 
+		if (this.canBeUsed())
+			this.initializationPromise = this.init();
+
 		this.setSettings = this.setSettings.bind(this);
 		this.setModel = this.setModel.bind(this);
 		this.getModels = this.getModels.bind(this);
 	}
 
+	private async init(): Promise<void> {
+		await this.ensureSession();
+		await Promise.all([
+			this.fetchVoices(),
+			this.fetchModels()
+		]);
+	}
+
 	/*
 		TTSModule methods
 	*/
@@ -79,7 +98,11 @@ export class ElevenLabsTTS implements TTSModule {
 	}
 
 	async generate(voice: string, text: string): Promise<TTSResponse> {
-		if (!this.voices) return {};
+		await this.initializationPromise;
+		await this.ensureSession();
+
+		if (!this.voices) return { error: 'no voices' };
+		if (!this.session) return { error: 'no session' };
 
 		const voiceData = this.voices.find((entry) => entry.name === voice);
 		if (!voiceData) return { error: 'Invalid voice' };
@@ -100,10 +123,8 @@ export class ElevenLabsTTS implements TTSModule {
 				'Sec-Fetch-Site': 'same-site',
 				'Sec-Fetch-Mode': 'cors',
 				'Sec-Fetch-Dest': 'empty',
-				host: 'api.us.elevenlabs.io',
-				...(config.tts_elevenlabs_token
-					? { Authorization: `Bearer ${config.tts_elevenlabs_token}` }
-					: { 'xi-api-key': config.tts_elevenlabs_key })
+				host: 'api.elevenlabs.io',
+				Authorization: `Bearer ${this.session.idToken}`
 			}
 		};
 
@@ -132,7 +153,7 @@ export class ElevenLabsTTS implements TTSModule {
 	}
 
 	canBeUsed(): boolean {
-		return config.tts_elevenlabs_key != undefined;
+		return config.tts_elevenlabs_refreshtoken != undefined && config.tts_elevenlabs_key != undefined;
 	}
 
 	/*
@@ -221,6 +242,35 @@ export class ElevenLabsTTS implements TTSModule {
 			});
 		});
 	}
+
+	private async ensureSession(): Promise<void> {
+		if (this.session && Date.now() < this.session.expiresAt - 300000)
+			return;
+
+		const refreshToken = this.session?.refreshToken || config.tts_elevenlabs_refreshtoken;
+		if (!refreshToken) throw new Error("No refresh token available");
+
+		const response = await fetch(FIREBASE_URL, {
+			method: 'POST',
+			headers: {
+				'Content-Type': 'application/x-www-form-urlencoded',
+				'Referer': 'https://elevenlabs.io/',
+				'Origin': 'https://elevenlabs.io'
+			},
+			body: new URLSearchParams({
+				grant_type: 'refresh_token',
+				refresh_token: refreshToken,
+			}),
+		});
+
+		if (!response.ok) throw new Error(`Auth Refresh Failed: ${await response.text()}`);
+		const data = await response.json();
+		this.session = {
+			idToken: data.id_token,
+			refreshToken: data.refresh_token,
+			expiresAt: Date.now() + (parseInt(data.expires_in) * 1000)
+		};
+	}
 }
 
 export default new ElevenLabsTTS();

src/utils/config.ts

@@ -7,7 +7,7 @@ export interface Config {
 	tts_default_voice: string | undefined;
 
 	tts_elevenlabs_key: string | undefined;
-	tts_elevenlabs_token: string | undefined;
+	tts_elevenlabs_refreshtoken: string | undefined;
 	tts_tiktok_sessionid: string | undefined;
 
 	steam_webapi_key: string | undefined;
@@ -31,7 +31,7 @@ function loadConfig(): Config {
 		tts_default_mode: process.env.DEFAULT_TTS_MODE,
 		tts_default_voice: process.env.DEFAULT_TTS_VOICE,
 		tts_elevenlabs_key: process.env.TTS_ELEVENLABS_KEY,
-		tts_elevenlabs_token: process.env.TTS_ELEVENLABS_TOKEN,
+		tts_elevenlabs_refreshtoken: process.env.TTS_ELEVENLABS_REFRESHTOKEN,
 		steam_webapi_key: process.env.STEAM_WEBAPI_KEY,
 		aws_access_id: process.env.AWS_ACCESS_ID,
 		aws_access_key: process.env.AWS_ACCESS_KEY,