Create app.py

app.py

@@ -0,0 +1,78 @@
+from flask import Flask, request, jsonify, send_file
+from flask_cors import CORS
+import torch
+import torchaudio
+import soundfile as sf
+import os
+from transformers import VitsModel, AutoTokenizer
+
+app = Flask(__name__)
+CORS(app)  # Allow Android requests
+
+# Model paths for different languages (Hugging Face Hub)
+MODELS = {
+    "kapampangan": "facebook/mms-tts-pam",
+    "tagalog": "facebook/mms-tts-tgl",
+    "english": "facebook/mms-tts-eng"
+}
+
+# Load models and processors
+loaded_models = {}
+loaded_processors = {}
+
+for lang, path in MODELS.items():
+    print(f"Loading {lang} model from Hugging Face: {path}...")
+    loaded_models[lang] = VitsModel.from_pretrained(path)
+    loaded_processors[lang] = AutoTokenizer.from_pretrained(path)
+    print(f"{lang.capitalize()} model loaded successfully!")
+
+SAMPLE_RATE = 16000  # Default sample rate
+OUTPUT_DIR = "/tmp/"  # Use /tmp for Hugging Face Spaces (limited storage)
+
+@app.route("/tts", methods=["POST"])
+def generate_tts():
+    """ API endpoint to generate speech based on the selected language. """
+    data = request.get_json()
+    text_input = data.get("text", "")
+    language = data.get("language", "kapampangan").lower()
+
+    if language not in MODELS:
+        return jsonify({"error": "Invalid language. Choose 'kapampangan', 'tagalog', or 'english'."}), 400
+
+    if not text_input:
+        return jsonify({"error": "No text provided"}), 400
+
+    print(f"Generating speech for: '{text_input}' in {language}")
+
+    # Select the correct model and processor
+    model = loaded_models[language]
+    processor = loaded_processors[language]
+
+    # Tokenize input text
+    inputs = processor(text_input, return_tensors="pt")
+
+    # Generate audio
+    with torch.no_grad():
+        output = model.generate(**inputs)
+
+    waveform = output.cpu().numpy().flatten()
+
+    # Save as WAV file
+    output_filename = os.path.join(OUTPUT_DIR, f"{language}_output.wav")
+    sf.write(output_filename, waveform, SAMPLE_RATE)
+
+    return jsonify({
+        "message": "TTS audio generated",
+        "file_url": f"/static/{language}_output.wav"
+    })
+
+@app.route("/static/<filename>")
+def serve_audio(filename):
+    """ Serve the generated WAV file. """
+    file_path = os.path.join(OUTPUT_DIR, filename)
+    if os.path.exists(file_path):
+        return send_file(file_path, mimetype="audio/wav")
+    return jsonify({"error": "File not found"}), 404
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=7860, debug=True)