vinay-ghate / Transcribe-api-with-whisper

API to get summary and timestamp from audio file

Audio Transcription and Summarization with FastAPI

This project provides a FastAPI-based server that handles audio files, transcribes them using the Whisper model, summarizes the content with the BART model, and extracts timestamps from the transcriptions.

Requirements

Install the required libraries:

pip install fastapi uvicorn whisper transformers torch python-dotenv

Usage

Start the Server

Run the FastAPI server using:

uvicorn api_application:app --reload

API Endpoint

POST /process-audio/

Upload an audio file to transcribe, summarize, and extract timestamps.

Request:

• Method: POST
• Content-Type: multipart/form-data
• Form Data:
  • file: The audio file to be processed (.wav, .mp3, etc.)

Response:

• transcript: The full transcription of the audio.
• summary: A summarized version of the transcription.
• timestamps: Extracted timestamps with corresponding transcript segments.

Example

You can use curl to test the endpoint:

curl -X POST "http://localhost:8000/process-audio/" -F "file=@/path/to/your/audiofile.mp3"

Project Structure

.
├── main.py          # Main FastAPI application
├── .env             # Environment variables (if needed)
├── README.md        # Project documentation