Living in an international environment, trying to communicate with people who speak different languages, the struggle is real.

_The problem: Having to switch between different apps for text input, translation, and voice output is a hassle we could all do without. I was wondering if I could leverage the power of LLMs to build something similar to a Google translate app with speech to text and also document translation.

_The solution: Enter PollyGlot - a unified translation platform that combines the power of OpenAI for natural translations, Whisper for speech recognition, and ElevenLabs for lifelike text-to-speech. Not just another translation tool; it’s my personal language bridge.

_Key features:

• Voice-to-text input using OpenAI’s Whisper
• Natural translations powered by GPT-4
• Lifelike text-to-speech with ElevenLabs
• Support for multiple languages (French, German, Spanish)
• File upload capabilities for text documents (better implementation for translating documents coming soon)
• Clean, intuitive interface built with shadcn/ui components

_Technical highlights: The project gave me hands-on experience with:

• React hooks for state management
• Next.js API routes for backend services
• Real-time voice processing
• Responsive design principles

_Learning outcomes: This project showed me how to work with multiple AI APIs in harmony, managing complex state in React applications and building accessible user interfaces.

Check out the code.

_Future plans: I’m working on adding support for PDF translation