Lisa-Shima/kinyarwanda-voice-ai-assistant--intelligent-robotics

# Kinyarwanda Voice Assistant A simple voice-enabled assistant that transcribes Kinyarwanda speech, matches it to predefined answers, and speaks the reply. Built with KinyaWhisper (ASR), fuzzy-wuzzy (NLP matching), and gTTS (TTS). --- ## Table of Contents - Features - Demo - Getting Started - Prerequisites - Installation - Usage - Project Structure - Customization --- ## Features - **ASR (Speech-to-Text):** Transcribes Kinyarwanda audio to text using the fine-tuned KinyaWhisper model. - **NLP Matching:** Matches transcribed questions to answers via fuzzy string matching. - **TTS (Text-to-Speech):** Converts answers into spoken audio with gTTS, falling back to English if Kinyarwanda unsupported. - **Multi-format Input:** Supports `.wav`, `.opus`, `.mp3` via FFmpeg/torchaudio. - **Sample Data:** Includes 5 example audio files and their expected responses. --- ## Demo > **Example Run:** > > ```bash > python main.py audio/q1_amakuru.wav > ``` > > Output: > > ``` > Transcribing… > You said: amakuru yawe? > Bot answers: Ni meza cyane, murakoze! > [warning] gTTS doesn’t support 'rw'. Falling back to English ('en'). > Generating speech… > Spoken reply saved to spoken_outputs/response_1682912345.mp3 > ``` --- ## Getting Started ### Prerequisites - Python 3.8 or newer - Git - FFmpeg (added to your system PATH) ### Installation 1. Clone the repo: ```bash git clone github.com cd kinyarwanda-voice-ai-assistant--intelligent-robotics ``` 2. Create and activate a virtual environment: ```bash python -m venv venv # Windows venv\Scripts\activate # macOS/Linux source venv/bin/activate ``` 3. Install dependencies: ```bash pip install -r requirements.txt ``` --- ## Usage 1. Run the full pipeline on a sample audio: ```bash python main.py audio/q1_amakuru.wav ``` 2. The assistant will: - Transcribe the speech (ASR) - Match to a predefined answer (NLP) - Speak the reply and save an MP3 (TTS) --- …