Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 from utils import generate_script, generate_audio, truncate_text, extract_text_from_url
 from prompts import SYSTEM_PROMPT
+from pydub import AudioSegment
 import pypdf
 import os
 import tempfile
@@ -29,21 +30,25 @@ def generate_podcast(file, url, tone, length):
         
         script = generate_script(SYSTEM_PROMPT, truncated_text, tone, length)
         
-        audio_files = []
+        audio_segments = []
         transcript = ""
-        for item in script.dialogue:
-            audio_file = generate_audio(item.text, item.speaker)
-            audio_files.append(audio_file)
-            transcript += f"**{item.speaker}**: {item.text}\n\n"
+        try:
+            for item in script.dialogue:
+                audio_file = generate_audio(item.text, item.speaker)
+                audio_segment = AudioSegment.from_mp3(audio_file)
+                audio_segments.append(audio_segment)
+                transcript += f"**{item.speaker}**: {item.text}\n\n"
+                os.remove(audio_file)  # Clean up temporary audio file
+        except Exception as e:
+            raise gr.Error(f"Error generating audio: {str(e)}")
         
-        # Combine audio files
-        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as combined_audio:
-            for audio_file in audio_files:
-                with open(audio_file, 'rb') as f:
-                    combined_audio.write(f.read())
-                os.remove(audio_file)  # Clean up individual audio files
+        combined_audio = sum(audio_segments)
         
-        return combined_audio.name, transcript
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
+            combined_audio.export(temp_audio.name, format="mp3")
+            temp_audio_path = temp_audio.name
+        
+        return temp_audio_path, transcript
     
     except Exception as e:
         return None, f"An error occurred: {str(e)}"
@@ -51,22 +56,25 @@ def generate_podcast(file, url, tone, length):
 instructions = """
 # Podcast Generator
 
-Welcome to the Podcast Generator project! This tool allows you to create custom podcast episodes using AI-generated content.
+Welcome to the Podcast Generator project! This tool creates custom podcast episodes using AI-generated content.
 
 ## Features
 * Generate podcast scripts from PDF content or web pages
 * Convert text to speech for a natural listening experience
-* Choose the tone of your podcast
+* Choose the tone of your podcast (Humorous, Casual, or Formal)
 * Export episodes as MP3 files
 
 ## How to Use
 1. Upload a PDF file OR enter a URL (content will be truncated to 2048 tokens if longer)
-2. Select the desired tone (humorous, casual, formal)
+2. Select the desired tone:
+   - Humorous: Expect jokes, puns, and playful banter
+   - Casual: Colloquial language, like a conversation between college students
+   - Formal: Professional podcast style with well-structured arguments
 3. Choose the podcast length
 4. Click "Generate" to create your podcast
 5. Listen to the generated audio and review the transcript
 
-Note: This tool uses the LLaMa 3.1 70B model for script generation and Voice RSS for text-to-speech conversion. The input is limited to 2048 tokens to ensure compatibility with the model. The podcast features John (Male, American accent) and Lily (Female, British accent) as hosts.
+Note: This tool uses the LLaMa 3.1 70B model for script generation and gTTS for text-to-speech conversion. The podcast features Sarah (American accent) and Maria (British accent) as hosts.
 """
 
 iface = gr.Interface(