wiameadnane/multilingual-speech-to-speech-translation-system

A multilingual speech-to-speech translation system that detects the input language using custom GMM models, transcribes, translates, and synthesizes speech. Supports English, French, Arabic, Spanish, Korean, and Moroccan Darija. # GlobalVoice: Multilingual Speech-to-Speech Translation System A complete end-to-end multilingual speech-to-speech translation platform that automatically detects the input language, transcribes speech to text, translates it to a target language, and synthesizes the translated text back to speech. ## Features - **Automatic Language Detection**: Uses custom-trained GMM (Gaussian Mixture Models) to identify the input language - **Speech Transcription**: Powered by OpenAI's Whisper model for accurate speech-to-text conversion - **Text Translation**: Utilizes Google Translate API for high-quality language translation - **Speech Synthesis**: Generates natural-sounding audio using Google Text-to-Speech (gTTS) - **User-Friendly GUI**: Modern PyQt6 interface with three intuitive screens - **Multi-language Support**: Supports 6 languages including English, French, Arabic, Spanish, Korean, and Moroccan Darija ## System Architecture The system follows a modular pipeline approach: ``` Input Audio → Language Detection (GMM) → Transcription (Whisper) → Translation (Google Translate) → Speech Synthesis (gTTS) → Output Audio ``` ### Core Components 1. **Language Detection Module** (`language_detection.py`) - Pre-trained GMM models with 512 components for 6-language detection - MFCC feature extraction with delta and delta-delta coefficients - WebRTC VAD for silence removal - Achieves 81.09% accuracy on 6 languages, 95% on 4 main languages 2. **Inference Pipeline** (`inference_pipeline.py`) - Whisper base model for transcription - Google Translate API integration - gTTS for speech synthesis - Audio processing and playback functionality 3. **User Interface** - **First Screen** (`first_screen.py`): Audio file upload - **Second Screen** (`second_screen.py`): Language detection and target language selection - **Third Screen** (`third_screen.py`): Results display with audio playback ## Supported Languages - **English** (en) - **French** (fr) - **Arabic** (ar) - **Spanish* …