/Audio-Segmentation-and-Analysis-for-call

Primary LanguageJupyter Notebook

Audio-Segmentation-and-Analysis-for-call

Audio Transcription with Whisper ASR

This code demonstrates how to transcribe audio segments using the OpenAI Whisper ASR model. It covers the following steps:

  1. Audio Download:

    • The code starts by downloading an audio file from Google Drive using its file ID.

  2. Audio Playback:

    • The downloaded audio file is played using IPython to verify its content.

  3. Audio Splitting:

    • The audio is split into two segments: caller and callee, each containing one half of the conversation.

  4. Whisper ASR Model:

    • The Whisper ASR model is loaded, and audio segments are transcribed using the model.

  5. VTT File Generation:

    • The transcriptions are saved in WebVTT (VTT) format, suitable for captioning and subtitles.

  6. Conversation Data Storage:

    • Transcriptions and audio segments' paths are stored in a JSON file for further analysis or use.

Prerequisites

Before running this code, make sure you have the necessary packages installed:

  • pydub
  • requests
  • openai
  • whisper

You can install these packages using pip:

pip install pydub requests openai whisper