uwimfuraz/KINYARWANDAVA

KinyaVA is a Kinyarwanda-speaking voice assistant that combines automatic speech recognition (ASR) and custom-trained text-to-speech (TTS) synthesis. Using NVIDIA NeMo and a VITS2 model, it enables voice interaction in Kinyarwanda, offering a foundation for AI applications in Rwanda and other low-resource regions. # Kinyarwanda_Voice_AI 🇷🇼 Kinyarwanda Voice Assistant - A locally-run AI assistant that understands and speaks Kinyarwanda. Features speech recognition (STT), natural language processing, and text-to-speech (TTS) with Gradio web interface. Ideal for Rwandan developers, researchers, and language preservation efforts. Demo Screenshot ## Features ✨ - 🎙️ Speech-to-text for Kinyarwanda using NeMo Conformer model - 🔊 Text-to-speech with custom KinyaTTS VITS2 model - 💬 Natural language processing for question answering - 🖥️ Gradio web interface with: - Microphone recording - Audio file upload - Automatic response playback - 🌐 Ready for deployment ## Installation 🛠️ ### Prerequisites - Python 3.8 or higher - pip package manager - (Optional) NVIDIA GPU for faster inference ### Setup 1. Clone the repository: ```bash git clone github.com cd kinyarwanda_Voice_AI ``` 2. Install dependencies: ```bash pip install -r requirements.txt ``` 3. Download models (automatic on first run): ```bash ASR: mbazaNLP/Kinyarwanda_nemo_stt_conformer_model TTS: Pre-trained KinyaTTS model ``` ### Usage 🚀 Running the Web Interface ```bash python interface.py ``` The interface will launch at: ```bash localhost ``` Using the Assistant Speak or upload audio in Kinyarwanda The assistant will: - Transcribe your speech - Generate an appropriate response - Speak back the response ### Project Structure 📂 ```bash . ├── interface.py # Gradio interface ├── assistant.py # Main assistant pipeline ├── stt_module.py # Speech-to-text processing ├── tts_module.py # Text-to-speech synthesis ├── nlp_module.py # Natural language processing ├── audio_samples/ # Example audio files ├── outputs/ # Generated response audio └── requirements.txt # Dependencies ``` ### Requirements ```bash gradio>=3.0 torch>=1.10 torchaudio>=0.10 nemo_toolkit[asr]>=1.7 soundfile>=0.10 huggingface_hub>=0.10 …