OpenAI Whisper - Local Test

This repository provides a simple setup and usage guide for transcribing audio files to text in English and Hindi using OpenAI's Whisper model.

Folder Structure

OpenAIwhisper-LocalTest/
├── README.md
├── install_venv.sh
├── requirements.txt
├── Dataset/
│   ├── Arzoooutput hindi.wav
│   └── Pankaj sir audio.wav
├── Research/
│   └── 1.Base.ipynb
└── output/
    ├── hinditranscription.txt
    └── transcription.txt

Installation

Step 1: Create and Activate Virtual Environment

# Create the virtual environment
python3 -m venv linuxvenv

# Activate virtual environment (Ubuntu/Linux)
source linuxvenv/bin/activate

# Upgrade pip
python -m pip install --upgrade pip

Step 2: Install Required Dependencies

pip install -r requirements.txt

Required Libraries

• JupyterLab
• OpenAI Whisper
• ffmpeg-python

Dataset

Audio files to be transcribed are placed in the Dataset directory:

• Arzoooutput hindi.wav (Hindi audio example)
• Pankaj sir audio.wav (English audio example)

Usage

Transcription using Jupyter Notebook

Open and run the provided notebook in the Research directory:

• 1.Base.ipynb

The notebook includes code blocks to:

• Transcribe English audio
• Transcribe Hindi audio

Running Transcription (CLI Example)

You can use Whisper directly from the command line as well:

whisper "Dataset/Pankaj sir audio.wav" --model base --language en
whisper "Dataset/Arzoooutput hindi.wav" --model medium --language hi

Output

Transcribed text files will be saved in the output directory:

• English transcription: transcription.txt
• Hindi transcription: hinditranscription.txt

Example Outputs

English Transcription:

Okay, next person is your supervisor...

Hindi Transcription:

मुझे अपनी आवाज विक्सित करने में काफी समय लगा और अब जब यह है तो मैं चुप नहीं रहने वाली

Notes

• Choose Whisper models based on required accuracy and performance:
  • tiny, base - Faster, less accurate
  • small, medium, large - More accurate, slower