app.py

from flask import Flask, request, jsonify

import whisper
import os
import tempfile
import io
import torchaudio

app = Flask(__name__)

# Initialize Whisper model
whisper_model = whisper.load_model("small")  # Renamed variable




@app.route('/transcribe', methods=['POST'])
def transcribe():
    try:
        # Read raw bytes from the request
        audio_bytes = request.data  
        if not audio_bytes:
            return jsonify({"error": "No audio data provided"}), 400

        # Convert bytes to a file-like object
        audio_file = io.BytesIO(audio_bytes)

        # Load audio as a waveform using torchaudio
        waveform, sample_rate = torchaudio.load(audio_file)

        # Whisper expects a NumPy array, so we convert it
        audio_numpy = waveform.squeeze().numpy()

        # Transcribe the audio
        result = model.transcribe(audio_numpy)

        return jsonify({"text": result["text"]})

    except Exception as e:
        print("Error:", str(e))  # Log error for debugging
        return jsonify({"error": "Internal Server Error", "details": str(e)}), 500