Update app.py

app.py

@@ -3,12 +3,27 @@ from TTS.api import TTS
 from TTS.tts.configs.vits_config import VitsConfig
 from huggingface_hub import hf_hub_download
 import os
-import json
-import gc 
+import gc
 from romanizer import sinhala_to_roman 
 
-def load_my_model(repo_id):
-    print(f"Downloading {repo_id}...")
+# Dictionary to hold our loaded models to avoid reloading every time
+models = {
+    "English": None,
+    "Sinhala": None,
+    "Tamil": None
+}
+
+def load_my_model(language):
+    # Mapping languages to their Hugging Face repos
+    repos = {
+        "English": "E-motionAssistant/text-to-speech-VITS-english",
+        "Sinhala": "E-motionAssistant/text-to-speech-VITS-sinhala",
+        "Tamil": "E-motionAssistant/text-to-speech-VITS-tamil"
+    }
+    
+    repo_id = repos[language]
+    
+    print(f"--- Loading {language} Model ---")
     model_path = hf_hub_download(repo_id=repo_id, filename="best_model.pth")
     config_path = hf_hub_download(repo_id=repo_id, filename="config.json")
 
@@ -16,57 +31,59 @@ def load_my_model(repo_id):
     config = VitsConfig()
     config.load_json(config_path)
 
-    if "english" in repo_id.lower():
-        print(f"--- SUCCESS: FORCING 137 CHARACTERS FOR ENGLISH ---")
+    # The English Character Fix (131 -> 137)
+    if language == "English":
+        print("Applying 137 character fix for English...")
         if hasattr(config, "model_args"):
             config.model_args.num_chars = 137
     
-    # Initialize TTS shell
+    # Initialize the TTS engine
     tts = TTS(gpu=False)
     
-    # Use positional arguments (no keywords) to avoid 'unexpected argument' errors
+    # Load using positional arguments for maximum compatibility
     tts.load_tts_model_by_path(model_path, config)
     
-    gc.collect()
+    gc.collect() # Clean up RAM
     return tts
 
-# --- Loading Models ---
-print("Initializing Models...")
-try:
-    eng_tts = load_my_model("E-motionAssistant/text-to-speech-VITS-english")
-    sin_tts = load_my_model("E-motionAssistant/text-to-speech-VITS-sinhala")
-    tam_tts = load_my_model("E-motionAssistant/text-to-speech-VITS-tamil")
-except Exception as e:
-    print(f"CRITICAL ERROR DURING LOADING: {e}")
-
 def generate_voice(text, language):
+    global models
     try:
-        if language == "English":
-            engine = eng_tts
-            processed_text = text
-        elif language == "Sinhala":
-            engine = sin_tts
-            # Your separate romanizer file logic
+        # Check if the model is already loaded, if not, load it now
+        if models[language] is None:
+            # Optional: Clear other models from RAM to stay under 16GB limit
+            # models = {k: None for k in models} 
+            # gc.collect()
+            models[language] = load_my_model(language)
+        
+        engine = models[language]
+        processed_text = text
+
+        # Apply Romanization logic for Sinhala
+        if language == "Sinhala":
             processed_text = sinhala_to_roman(text)
-        else:
-            engine = tam_tts
-            processed_text = text
+            print(f"Sinhala Romanized: {processed_text}")
 
-        output_path = "output.wav"
-        engine.tts_to_file(text=processed_text, file_path=output_path)
+        # Generate the audio file
+        output_path = f"output_{language.lower()}.wav"
+        engine.tts_to_file(text=str(processed_text), file_path=output_path)
+        
         return output_path
+
     except Exception as e:
-        return f"Error: {e}"
+        print(f"Error during {language} generation: {e}")
+        return None
 
 # Gradio Interface
 demo = gr.Interface(
     fn=generate_voice,
     inputs=[
-        gr.Textbox(label="Input Text"),
-        gr.Dropdown(["English", "Sinhala", "Tamil"], label="Select Language")
+        gr.Textbox(label="Input Text", placeholder="Type your message here..."),
+        gr.Dropdown(["English", "Sinhala", "Tamil"], label="Select Language", value="English")
     ],
-    outputs=gr.Audio(label="Synthesized Speech", type="filepath"),
-    title="Multilingual VITS TTS"
+    outputs=gr.Audio(label="Generated Speech", type="filepath"),
+    title="Multilingual VITS TTS System",
+    description="A lightweight Text-to-Speech system for English, Sinhala, and Tamil."
 )
 
 if __name__ == "__main__":