app.py

import os
import numpy as np
from flask import Flask, request, jsonify, send_file, send_from_directory
import google.generativeai as genai
from gtts import gTTS, lang
import tempfile
import soundfile as sf
from kokoro import KPipeline
from werkzeug.utils import secure_filename
from flask_cors import CORS

app = Flask(__name__, static_folder='static')
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
KOKORO_LANGUAGES = {
    "American English": "a",
    "British English": "b",
    "Mandarin Chinese": "z",
    "Spanish": "e",
    "French": "f",
    "Hindi": "h",
    "Italian": "i",
    "Brazilian Portuguese": "p"
}

GTTS_LANGUAGES = lang.tts_langs()
GTTS_LANGUAGES['ja'] = 'Japanese'  # Explicit Japanese support

SUPPORTED_LANGUAGES = sorted(
    list(set(list(KOKORO_LANGUAGES.keys()) + list(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

        # Validate MIME type
        allowed_mime_types = ['audio/wav', 'audio/mpeg', 'audio/mp4', 'audio/webm']
        if audio_file.mimetype not in allowed_mime_types:
            return jsonify({'error': f'Unsupported file type: {audio_file.mimetype}'}), 400

        # Transcribe audio using Gemini
        model = genai.GenerativeModel("gemini-2.0-flash")
        
        # Create proper audio blob
        audio_blob = {
            'mime_type': audio_file.mimetype,
            'data': audio_file.read()
        }

        # Get transcription
        convo = model.start_chat()
        convo.send_message("You are a professional transcriber. Transcribe this audio accurately and verbatim in the original language. Respond only with the transcription.")
        response = convo.send_message(audio_blob)
        transcription = response.text.strip()

        # Translate text using Gemini
        prompt = f"Translate the following text to {target_language} preserving meaning and cultural nuances. Respond only with the translation:\n\n{transcription}"
        response = model.generate_content(prompt)
        translated_text = response.text.strip()
        
        # Generate TTS
        if target_language in KOKORO_LANGUAGES:
            lang_code = KOKORO_LANGUAGES[target_language]
            pipeline = KPipeline(lang_code=lang_code)
            generator = pipeline(translated_text, voice="af_heart", speed=1)
            
            # Collect all audio segments
            audio_segments = []
            for _, _, audio in generator:
                if audio is not None:
                    audio_segments.append(audio)
            
            if audio_segments:
                audio_data = np.concatenate(audio_segments)
                _, temp_output_path = tempfile.mkstemp(suffix=".wav")
                sf.write(temp_output_path, audio_data, 24000)
            else:
                raise ValueError("No audio generated by Kokoro")
        else:
            # Standard gTTS handling
            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:
        app.logger.error(f"Error processing request: {str(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=7860)