/VN-ASR-Demo

Primary LanguagePython

Vietnamese Speech to Text Web Application

Overview

This web application is designed for Vietnamese Speech to Text conversion, allowing users to transcribe audio files or speak directly into the microphone. The application is built using Streamlit and utilizes the powerful Wav2Vec 2.0 model for speech recognition. The model has been fine-tuned from the pre-trained nguyenvulebinh/wav2vec2-base-vietnamese-250h on a dataset obtained from the BKAI2023 competition.

Features

  • Audio Input Options:

    • Import audio files in .wav or .mp3 format.
    • Directly speak into the microphone for real-time transcription.

  • Wav2Vec 2.0 Model:

    • The speech recognition is powered by the Wav2Vec 2.0 model, known for its accuracy in transcribing spoken language.

  • Fine-Tuned Model:

Getting Started

Prerequisites

  • Python 3.7 or higher
  • Install required packages: pip install -r requirements.txt

Running the Application

  1. Clone the repository: git clone https://github.com/trongnk2106/VN-ASR-Demo.git
  2. Navigate to the project directory: cd VN-ASR-Demo
  3. Run the application: streamlit run app.py
  4. Access the web application at http://localhost:8501 in your web browser.

Usage

  1. File Input:

    • Click on the "Upload Audio File" button and select a .wav or .mp3 file for transcription.

  2. Microphone Input:

    • Click on the "Start Microphone" button and speak into your microphone for real-time transcription.

  3. View Transcription:

    • The transcribed text will be displayed on the web page.

Acknowledgments

License

This project is licensed under the MIT License - see the LICENSE file for details.

Issues and Contributions

If you encounter any issues or would like to contribute to the project, please open an issue or pull request on the GitHub repository.