Talk2Viet AI - Real-Time Voice Dubbing

Talk2Viet AI dubs any YouTube video into Vietnamese — live, while you watch. WHAT IT DOES - Detects the video's subtitle track (real or auto-generated). - Translates segments into natural Vietnamese using a domain-aware AI prompt (news, gaming, education, tech, lifestyle, sports, kids, vlog, and more). - Reads the translation aloud with Microsoft Edge's neural TTS voices. - Streams chunks while you watch — first audio plays within ~5 seconds for typical videos. HOW TO USE 1. Click "Sign in with Google" in the extension popup (one-time). 2. Open any YouTube video that has captions. 3. Click "Listen" in the floating control. Vietnamese audio starts. 4. Scrub the video — translation re-prioritises around your current time. WHY IT'S DIFFERENT - Domain-aware prompts: the LLM classifies the video's topic first, then translates with vocabulary appropriate to that domain (e.g. it won't translate "deploy" or "API" inside tech videos). - Shared cache: once one user has translated a video, every subsequent viewer gets the result instantly. Your videos warm the cache for others. - Stream-first architecture: chunks render as they're ready, not after the whole video has been translated. PRIVACY We only collect what's necessary to operate the service: your Google email + display name (for sign-in), a session token, and the subtitle text of the videos you ask us to translate. We do not track your watch history, browsing data, or anything outside the YouTube pages where you actively use the extension. Full privacy policy: https://talk2viet.vercel.app/privacy UNDER THE HOOD - Auth: Firebase Google sign-in. - Backend: Hono on Vercel, Neon Postgres, Upstash Redis. - Translation: OpenRouter-routed LLM (Google Gemini Flash) with carefully tuned domain prompts. - TTS: Microsoft Edge neural voices. Built for Vietnamese viewers who want to learn from English (and other-language) creators without waiting for human dubs or fan-subs.