EngVision Whisper Service

Overview

EngVision Whisper is a speech-to-text service leveraging the Whisper model from OpenAI. It's designed to transcribe audio files efficiently and accurately. The service is built using Flask and can be easily deployed in a containerized environment.

Requirements

• Docker
• Python 3.7 or later

Installation and Setup

Building the Docker Image

1. Clone the Repository: Clone the repository to your local machine.

   git clone https://github.com/EngVision/Whisper-Service
   cd Whisper-Service
   

2. Build Docker Image:

   docker build -t engvision-whisper .
   

Running the Container

Run the Docker container:

docker run -p 8000:8000 engvision-whisper

Usage

Starting the Service

Once the container is running, the service will be available at http://localhost:8000/.

API Endpoints

1. Root Endpoint (GET /):

   • Description: A simple endpoint to confirm that the service is running.
   • Response: "Hello from EngVision Whisper!"

2. Speech to Text Endpoint (POST /stt/<file_id>):

   • Description: Upload an audio file for transcription.
   • Response:

   {
     "_id": "65dac9d6b407958caa1af7f7",
     "file_id": "65dac9d6b407958caa1af7f7",
     "status": "processing",
     "text": null
   }

3. Check Status Endpoint (GET /stt/<file_id>):

   • Description: Check the status of a transcription task.
   • Response: A JSON with the status of the task and the result if completed.

   {
     "_id": "65dac9d6b407958caa1af7f7",
     "file_id": "65dac9d6b407958caa1af7f7",
     "status": "completed",
     "text": " This is my favorite food."
   }

4. Speech Evaluation Endpoint (POST /speech-evaluation):

   • Description: Get IPA and evaluation.
   • Body:

   {
     "fileId": "65dac87bd469ca2adaed6f98",
     "original": "This is my favourite food"
   }

   • Response:

   {
     "_id": "65dac87bd469ca2adaed6f98",
     "correct_letters": "111 11 111 111111111 111 ",
     "original_ipa_transcript": "ðɪs ɪz maɪ ˈfeɪvərɪt fud",
     "original_transcript": "This is my favourite food",
     "pronunciation_accuracy": "95",
     "submission_id": "65dac87bd469ca2adaed6f98",
     "voice_ipa_transcript": "ðɪs ɪz maɪ ˈfeɪvərɪt fud",
     "voice_transcript": "this is my favorite food"
   }

Example Usage

1. Uploading an Audio File for Transcription:

   curl -F "file=@path_to_audio_file" http://localhost:8000/stt
   

2. Checking Transcription Status:

   curl http://localhost:8000/stt/<file_id>
   

Contributing

Contributions are welcome. Please fork the repository and submit a pull request with your changes.