app.py

import os
import base64
from flask import Flask, request, jsonify, send_file, send_from_directory
import google.generativeai as genai
from gtts import gTTS, lang
import tempfile
from werkzeug.utils import secure_filename
from flask_cors import CORS

app = Flask(__name__, static_folder='static', static_url_path='')
CORS(app)

# Configure Gemini API
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
if not GEMINI_API_KEY:
    raise ValueError("GEMINI_API_KEY environment variable not set")
genai.configure(api_key=GEMINI_API_KEY)

# Language configurations
GTTS_LANGUAGES = lang.tts_langs()
SUPPORTED_LANGUAGES = sorted(GTTS_LANGUAGES.values())

@app.route('/')
def serve_index():
    return send_from_directory(app.static_folder, 'index.html')

@app.route('/languages')
def get_languages():
    return jsonify(SUPPORTED_LANGUAGES)

@app.route('/translate', methods=['POST'])
def translate_audio():
    try:
        if 'audio' not in request.files:
            return jsonify({'error': 'No audio file uploaded'}), 400
            
        audio_file = request.files['audio']
        target_language = request.form.get('language', 'English')
        
        if not audio_file or audio_file.filename == '':
            return jsonify({'error': 'Invalid audio file'}), 400

        # Save temporary audio file
        filename = secure_filename(audio_file.filename)
        temp_input_path = os.path.join(tempfile.gettempdir(), filename)
        audio_file.save(temp_input_path)
        
        # Read audio file as base64
        with open(temp_input_path, "rb") as f:
            audio_data = base64.b64encode(f.read()).decode("utf-8")
        
        # Transcribe with Gemini
        model = genai.GenerativeModel("gemini-1.5-pro-latest")
        prompt = """Accurately transcribe this audio file. Return only the raw text without formatting."""
        
        response = model.generate_content(
            [
                prompt,
                {
                    "mime_type": "audio/" + filename.split('.')[-1],
                    "data": audio_data
                }
            ]
        )
        transcription = response.text.strip()
        
        # Translate with Gemini
        translate_prompt = f"Translate to {target_language} preserving meaning: {transcription}"
        translated_response = model.generate_content(translate_prompt)
        translated_text = translated_response.text.strip()
        
        # Generate TTS
        lang_code = next((k for k, v in GTTS_LANGUAGES.items() if v == target_language), 'en')
        tts = gTTS(translated_text, lang=lang_code)
        _, temp_output_path = tempfile.mkstemp(suffix=".mp3")
        tts.save(temp_output_path)
        
        return jsonify({
            'transcription': transcription,
            'translation': translated_text,
            'audio_url': f'/download/{os.path.basename(temp_output_path)}'
        })
        
    except Exception as e:
        return jsonify({'error': str(e)}), 500

@app.route('/download/<filename>')
def download_file(filename):
    try:
        return send_file(
            os.path.join(tempfile.gettempdir(), filename),
            mimetype="audio/mpeg",
            as_attachment=True,
            download_name=f"translated_{filename}"
        )
    except FileNotFoundError:
        return jsonify({'error': 'File not found'}), 404

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=int(os.environ.get('PORT', 5000)))