Update app.py

app.py

@@ -2,18 +2,25 @@ import gradio as gr
 from supertonic import TTS
 import os
 import numpy as np
+import tempfile
+import spaces  # ZeroGPU के लिए अनिवार्य
+import torch
+
+# GPU डिटेक्ट करना और TTS को GPU पर लोड करना
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# Initialize TTS - auto_download=True handles the HF model fetching automatically
-# This runs on CPU by default via ONNX
 try:
+    # अगर Supertonic लाइब्रेरी device पैरामीटर सपोर्ट करती है
+    tts = TTS(auto_download=True, device=device)
+except TypeError:
+    # अगर यह डिफ़ॉल्ट रूप से ONNX का उपयोग कर रही है
     tts = TTS(auto_download=True)
+    print("Warning: Initialization using default backend.")
 except Exception as e:
     print(f"Error initializing TTS: {e}")
 
-# All available voices found in the repository tree
 VOICES = ["M1", "M2", "M3", "M4", "M5", "F1", "F2", "F3", "F4", "F5"]
 
-# The 31 supported languages
 LANGUAGES = {
     "English": "en", "Korean": "ko", "Japanese": "ja", "Arabic": "ar",
     "Bulgarian": "bg", "Czech": "cs", "Danish": "da", "German": "de",
@@ -25,6 +32,7 @@ LANGUAGES = {
     "Turkish": "tr", "Ukrainian": "uk", "Vietnamese": "vi"
 }
 
+@spaces.GPU  # यह लाइन जनरेशन के समय ZeroGPU को एक्टिवेट करेगी
 def generate_speech(text, voice, language_name):
     if not text.strip():
         raise gr.Error("Please enter some text.")
@@ -32,17 +40,15 @@ def generate_speech(text, voice, language_name):
     try:
         lang_code = LANGUAGES[language_name]
         
-        # Get the voice style object
         style = tts.get_voice_style(voice_name=voice)
-        
-        # Synthesize (Returns wav data and a numpy array for duration)
         wav, duration = tts.synthesize(text, voice_style=style, lang=lang_code)
         
-        # Save to a temporary path
-        output_path = "output.wav"
+        # ERROR FIX: हर ऑडियो के लिए एक यूनीक टेम्परेरी फाइल बनाना
+        temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
+        output_path = temp_file.name
+        
         tts.save_audio(wav, output_path)
         
-        # FIX: Convert numpy.ndarray duration to float for f-string compatibility
         readable_duration = float(duration)
         
         return output_path, f"Generation Successful! \nDuration: {readable_duration:.2f}s"
@@ -50,10 +56,10 @@ def generate_speech(text, voice, language_name):
     except Exception as e:
         raise gr.Error(f"Generation failed: {str(e)}")
 
-# Define the Gradio Interface
-with gr.Blocks(theme='soft', title="Supertonic 3 TTS") as demo:
+# Gradio Interface
+with gr.Blocks(theme='soft', title="Supertonic 3 TTS (GPU Accelerated)") as demo:
     gr.Markdown("# 🎙️ Supertonic 3: Multilingual TTS")
-    gr.Markdown("An on-device, lightweight Text-to-Speech system by **Supertone**. Running on CPU via ONNX.")
+    gr.Markdown("An on-device, lightweight Text-to-Speech system by **Supertone**. Accelerated via **ZeroGPU**.")
     
     with gr.Row():
         with gr.Column(scale=1):