Logo Lanfrica

Israel-Taddese/amharic-voice-ai

Domaine:

natural language processing

Type de record:

software
Créateur:
Isr
Hôte:
Azure-powered Amharic ↔ English speech and text translation app with FastAPI, Azure Speech, Azure Translator, and local web UI. # AmharicVoice AI ## Portfolio Project Summary **AmharicVoice AI** is an Azure-powered Amharic ↔ English translation MVP that supports typed text translation, browser microphone recording, WAV speech translation, translated audio playback, and Amharic-aware phrase normalization. This project demonstrates full-stack application development, cloud API integration, secure backend design, deployment documentation, and practical troubleshooting across browser audio, Azure AI services, FastAPI, Render, GitHub, and Windows development environments. ## Live Demo - **Frontend demo:** `amharic-voice-ai-web.onrend…` - **Backend health check:** `amharic-voice-ai.onrender.c…` - **Backend API docs:** `amharic-voice-ai.onrender.c…` > Note: The backend is hosted on Render's free tier, so the first request after inactivity may take extra time while the service wakes up. ## Key Features - Amharic ↔ English text translation - Amharic ↔ English speech translation - Browser microphone recording - Browser recordings encoded as 16 kHz mono 16-bit PCM WAV - Direct upload of recorded WAV audio to the speech translation endpoint - Translated audio playback - Custom Amharic phrase normalization for greeting phrases and speech-recognition variations - Public frontend and backend deployment on Render - API documentation through FastAPI Swagger UI ## Security and Privacy Highlights - `.env` file excluded from Git tracking - `.env.example` included for safe configuration reference - Azure keys stored as Render environment variables - Rate limiting added to reduce API abuse risk - Upload file size limits added for speech requests - Chunked upload validation added to avoid reading oversized uploads into memory - Invalid WAV error handling added for renamed `.m4a`, `.mp3`, or compressed audio files - Generated translated audio cleanup added - Production CORS restricted to the deployed frontend - Privacy/security checklist documented ## Technica …

Languages