Update app.py

app.py

@@ -75,12 +75,8 @@ def _init_pocket(
     import os
     hf_token = os.environ.get("HF_TOKEN")
     if hf_token:
-        try:
-            from huggingface_hub import login
-            login(token=hf_token, add_to_git_credential=False)
-            print("Logged in to HuggingFace")
-        except Exception as e:
-            print(f"HuggingFace login failed: {e}")
+        print("HF_TOKEN found, using for authentication")
+
 
     # Auto-detect device: CPU by default, CUDA if available
     # Note: The pocket-tts docs mention GPU doesn't provide speedup for this model
@@ -276,6 +272,13 @@ def _get_voice_state(voice_name: str | None, custom_audio_path: str | None):
         return voice_state
 
     # Fall back to voice cloning (requires auth)
+    if not model.has_voice_cloning:
+        raise gr.Error(
+            f"No embedding found for voice '{voice_name}'. "
+            f"Voice cloning is not available (requires HF auth). "
+            f"Run the app locally first to create embeddings."
+        )
+    
     voice_path = PRESET_VOICES[voice_name]
     print(f"Loading preset voice '{voice_name}' from: {voice_path}")
     
@@ -432,8 +435,17 @@ def generate_tts(
 with gr.Blocks() as demo:
     gr.HTML(
         "<h1 style='text-align: center;'>Pocket-TTS</h1>"
-        "<p style='text-align: center;'>Powered by kyutai/pocket-tts | Lightweight TTS on CPU</p>"
     )
+    device_info = gr.Markdown(
+        "<p style='text-align: center;'>Powered by kyutai/pocket-tts | Running on CPU</p>"
+    )
+    
+    @demo.load
+    def update_device_info():
+        device = "CUDA" if torch.cuda.is_available() else "CPU"
+        return f"<p style='text-align: center;'>Powered by kyutai/pocket-tts | Running on {device}</p>"
+    
+    demo.load(update_device_info, outputs=device_info)
 
     with gr.Row():
         with gr.Column():