Update app.py

app.py

@@ -1,49 +1,47 @@
-import os
 import gradio as gr
-from TTS.api import TTS
-import noisereduce as nr
+from transformers import pipeline
 import soundfile as sf
+import torch
+from datasets import load_dataset
 import gc
 
-# Memory clear karo taaki server hamesha zinda rahe
+# System ki purani memory saaf karna
 gc.collect()
-os.environ['COQUI_TOS_AGREED'] = '1'
 
-print("⏳ Loading VoiceForge Cloning Engine...")
-# Sabse stable cloning engine jo HF free tier par nahi rukta
-tts = TTS("tts_models/multilingual/multi-dataset/your_tts")
-print("✅ Engine Ready & Running!")
+print("⏳ WARNING: Loading the BIGGEST Engine... Server crash hone ke chances hain!")
 
-def generate_api_voice(text, reference_audio):
-    if not text or not reference_audio:
-        return None, "Error: Script aur Reference Voice dono zaroori hain!"
+try:
+    # Heavy Text-to-Speech pipeline load kar rahe hain
+    synthesizer = pipeline("text-to-speech", "microsoft/speecht5_tts")
     
-    try:
-        # Background noise saaf karne ka automatic system
-        data, rate = sf.read(reference_audio)
-        if len(data.shape) > 1: data = data.mean(axis=1)
-        clean_data = nr.reduce_noise(y=data, sr=rate)
-        sf.write("clean_ref.wav", clean_data, rate)
+    # High-quality speaker embedding
+    embeddings_dataset = load_dataset("Matthijs/cmu_arctic_xvectors", split="validation")
+    speaker_embedding = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
+    print("✅ Engine load ho gaya! (Ye ek miracle hai)")
+except Exception as e:
+    print(f"❌ Server Crashed: {e}")
 
+def generate_api_voice(text):
+    if not text:
+        return None, "Error: Script likhna zaruri hai!"
+    
+    try:
+        # Aawaz generate karne ka heavy process
+        speech = synthesizer(text, forward_params={"speaker_embeddings": speaker_embedding})
+        
         output_file = "output_voice.wav"
-        tts.tts_to_file(text=text, speaker_wav="clean_ref.wav", language="en", file_path=output_file)
+        sf.write(output_file, speech["audio"], samplerate=speech["sampling_rate"])
         
         return output_file, "✅ API Status: Success"
     except Exception as e:
         return None, f"⚠️ Engine Error: {str(e)}"
 
-# Ekdum clean UI (Bina kisi extra pins ya technical labels ke)
+# Ekdum clean UI (No extra clutter)
 iface = gr.Interface(
     fn=generate_api_voice,
-    inputs=[
-        gr.Textbox(label="Apni Script Yahan Likho"), 
-        gr.Audio(type="filepath", label="Reference Voice Upload")
-    ],
-    outputs=[
-        gr.Audio(label="VoiceForge Output"), 
-        gr.Textbox(label="Status")
-    ],
-    title="🎙️ VoiceForge AI Studio"
+    inputs=[gr.Textbox(label="Apni Script Yahan Likho")],
+    outputs=[gr.Audio(label="VoiceForge Output"), gr.Textbox(label="Status")],
+    title="🎙️ VoiceForge AI Studio - Heavy Engine Test"
 )
 
 iface.launch()