MICHI — AI Voice Translator

A real-time, two-person AI voice translator built for natural face-to-face conversations. Collaborated with Jhames

---

Features

• Voice-Powered Translation – Hold to speak, release to translate
• Split-Screen Interface – Two speakers, one conversation
• 70+ Languages – Including Filipino, Cebuano, and regional languages
• Context-Aware AI – Remembers the last 5 exchanges for better translations
• Mobile-Optimized – Upside-down mode for two-person use on phones
• Secure API Calls – API keys managed via Netlify serverless functions
• Keyboard Shortcuts – Press Q (Speaker 1) or P (Speaker 2) on desktop

---

How It Works

1. Both Speakers selects their language
2. Speaker 1 holds the mic button and speak
3. Speaker 2 sees the translation appear on their side
4. Speaker 2 responds — and the conversation flows naturally
5. The AI remembers context for more accurate translations

Conversation Memory: The AI uses the last 5 exchanges to translate better. No more awkward "it" or "this" translations!

---

Tech Stack

Layer	Technology
Frontend	HTML + CSS + Vanilla JS
UI Style	Neumorphic Design
Speech-to-Text	Web Speech API
Translation AI	Groq API (Llama 3.3)
Backend	Netlify Serverless Functions
Hosting	Netlify
Version Control	GitHub

---

Click here to try MICHI