benax-rw/SpeechBrain-with-Kinyarwanda

SpeechBrain-with-Kinyarwanda is a project for building and training Kinyarwanda ASR models using SpeechBrain. It includes datasets, training scripts, models, and tools to support speech recognition research and applications in Kinyarwanda. # 🧠 SpeechBrain-with-Kinyarwanda A lightweight Kinyarwanda automatic speech recognition (ASR) project built with SpeechBrain. This repo contains code for converting audio, transcribing speech, and experimenting with pretrained models for Kinyarwanda. --- ## 📥 Clone the Project ```bash git clone github.com cd SpeechBrain-with-Kinyarwanda ``` --- ## 🧪 Set Up the Environment You can (optionally) use a virtual environment: ```bash python3 -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows ``` --- ## 📦 Install Dependencies ```bash pip install -r requirements.txt # if available ``` Or manually: ```bash pip install speechbrain torchaudio ``` --- ## 🎤 Run a Transcription The model will **automatically download on first run**. ```bash python transcribe_kinyarwanda.py rw-103.wav ``` --- ## 🛠 Tools Included - `convert_m4a_to_mp3.py` – convert `.m4a` files to `.mp3` - `convert_m4a_to_wav.py` – convert `.m4a` to `.wav` for ASR - `transcribe_kinyarwanda.py` – runs inference using pretrained model - Sample files: `rw-103.wav`, `rw-104.wav`, etc. --- ## 🙌 Acknowledgements - Built with ❤️ using SpeechBrain - Kinyarwanda models from Hugging Face --- ## 📚 License MIT – free for educational use and research.