Update app.py

app.py

@@ -3,25 +3,22 @@ import gradio as gr
 import requests
 import uuid
 import os
-import torch
 import soundfile as sf
 from f5_tts.api import F5TTS
 
-# 1. Configuration & Device Auto-Detection
+# ==========================================
+# 1. Setup Configuration & Model Loading
+# ==========================================
 GROQ_API_KEY = "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"
 voice_file = "my_voice_fixed.wav"
 
-# Automatically detect if space has access to GPU acceleration or fallback to CPU
-device = "cuda" if torch.cuda.is_available() else "cpu"
-print(f"🚀 Initializing simulation engine on device: {device}")
-
 print("Loading Whisper Model...")
-whisper_model = whisper.load_model("base", device=device)
+whisper_model = whisper.load_model("base")
 
-print("Loading F5-TTS Simulation Model...")
-f5tts = F5TTS(device=device)
+print("Loading F5-TTS Core Simulation Model...")
+f5tts = F5TTS()
 
-# Automatically extract reference text from the file at startup
+# Automatically extract reference text from your voice file at startup
 print(f"Extracting reference text from {voice_file}...")
 try:
     ref_result = whisper_model.transcribe(voice_file)
@@ -33,7 +30,9 @@ except Exception as e:
     print(f"⚠️ Text extraction warning: {e}")
     auto_reference_text = "Hello"
 
-# 2. Resilient API Function with Smart Text Fallback
+# ==========================================
+# 2. Text Response Function (Groq API)
+# ==========================================
 def get_ai_answer(question):
     try:
         response = requests.post(
@@ -50,44 +49,52 @@ def get_ai_answer(question):
             timeout=10
         )
         if response.status_code == 200:
+            # Fixed list indexing to unpack payload properly
             return response.json()["choices"][0]["message"]["content"].strip()
     except Exception as e:
-        print(f"Groq API connection down: {e}")
+        print(f"Groq API connection error: {e}")
     
-    # 🛠️ SMART FALLBACK: If API fails, supply valid text so voice cloning NEVER breaks
+    # Text fallback baseline to prevent downstream TTS pipeline crashes
     q_lower = question.lower()
     if "salesforce" in q_lower:
         return "Salesforce is a cloud-based customer relationship management platform that helps businesses manage sales, marketing, and customer service tracking in one place."
-    return "Hello! I am your voice simulation assistant. Your system is online and ready for audio cloning."
+    return "Hello! Your voice simulation system is online and ready for audio cloning."
 
-# 3. Robust Voice Simulation Block
+# ==========================================
+# 3. Voice Generation Block (F5-TTS)
+# ==========================================
 def clone_and_speak(text):
     try:
         unique_filename = f"voice_{uuid.uuid4().hex[:8]}.wav"
-        print(f"🔊 Generating cloned speech for text: '{text}' -> Saving to: {unique_filename}")
+        print(f"🔊 Initiating F5-TTS generation -> Targeting: {unique_filename}")
         
-        # Execute voice simulation inference
+        # Generates and writes the file directly to your workspace using the unique name
         f5tts.infer(
             ref_file=voice_file,
             ref_text=auto_reference_text,
             gen_text=text,
-            file_wave=unique_filename
+            file_wave=unique_filename,
+            nfe_step=32, # Balanced steps for clean sound quality output
+            speed=1.0
         )
         
-        # Format the audio file structure to enforce strict Gradio/Browser playback support
+        # Verify the file exists on disk and is formatted correctly for your browser player
         if os.path.exists(unique_filename) and os.path.getsize(unique_filename) > 0:
+            print(f"🎉 Voice generation successful! File written to disk: {unique_filename}")
             data, samplerate = sf.read(unique_filename)
             sf.write(unique_filename, data, 24000, format='WAV', subtype='PCM_16')
-            print(f"🎉 Cloned audio track generated successfully: {unique_filename}")
             return unique_filename
         else:
-            print("❌ Audio file generation mismatch: File is empty.")
+            print("❌ Audio generation failed: File was not created on disk.")
             return None
             
     except Exception as e:
-        print(f"❌ F5-TTS Voice Inference Error Details: {e}")
+        print(f"❌ F5-TTS Pipeline Exception Breakdown: {e}")
         return None
 
+# ==========================================
+# 4. Main Chatbot Routing & Logic
+# ==========================================
 def chatbot(text_input, audio_input):
     if audio_input is not None:
         result = whisper_model.transcribe(audio_input)
@@ -101,7 +108,9 @@ def chatbot(text_input, audio_input):
     audio_path = clone_and_speak(answer)
     return answer, audio_path
 
-# 4. Gradio Interface Construction
+# ==========================================
+# 5. Gradio Interface Construction
+# ==========================================
 interface = gr.Interface(
     fn=chatbot,
     inputs=[