This Jupyter Notebook demonstrates how to perform real-time speech transcription using the Whisper model from Hugging Face's Transformers library. The notebook captures audio from the microphone, processes it, and prints the transcription in real-time.
To get started, clone the repository using the following command:
git clone https://github.com/Viraj-08/Marathi-Speech-To-Text.gitNavigate to the cloned directory and set up a virtual environment:
cd Marathi-Speech-To-Text
python -m venv .venv
.venv\Scripts\activate # On Windows
# source .venv/bin/activate # On macOS/LinuxInstall the required packages by running:
pip install -r requirements.txtFor PyTorch, install it using the following command:
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118Run the code in the Jupyter Notebook using the above virtual environment.
- This notebook uses the Whisper small model fine-tuned on the Marathi language. You can replace it with any other Whisper model available on the Hugging Face Hub.