Update app.py

app.py

@@ -508,6 +508,30 @@ with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
     # =============================
     # Voice Chat Functions
     # =============================
+    def text_to_speech(text):
+        # Convert text to speech using Bark
+        from transformers import AutoProcessor, BarkModel
+        import numpy as np
+    
+        processor = AutoProcessor.from_pretrained("suno/bark-small")
+        model = BarkModel.from_pretrained("suno/bark-small")
+    
+        inputs = processor(text, voice_preset="v2/en_speaker_6", return_tensors="pt")
+        speech = model.generate(**inputs)
+    
+        # ✅ Extract and normalize audio data
+        audio_data = speech["audio"]
+        sampling_rate = speech["sampling_rate"]
+    
+        # 🔊 Normalize & clip Bark audio output to avoid struct.error
+        if isinstance(audio_data, np.ndarray):
+            audio_data = np.clip(audio_data, -1.0, 1.0).astype(np.float32)
+        else:
+            audio_data = np.array(audio_data, dtype=np.float32)
+            audio_data = np.clip(audio_data, -1.0, 1.0)
+    
+        return (sampling_rate, audio_data)
+
     def process_voice_input(audio, history, temp, max_tok, topk, session_id):
         """Process voice input: transcribe, get response, convert to speech"""
         if audio is None: