Update app.py

app.py

@@ -1,40 +1,90 @@
 import gradio as gr
-from utils import generate_script, generate_audio
+from utils import generate_script, generate_audio, truncate_text
 from prompts import SYSTEM_PROMPT
 from pydub import AudioSegment
+import pypdf
 import os
+import io
+import tempfile
 
 def generate_podcast(file, tone, length):
     # Extract text from PDF
-    # Generate script using generate_script function
+    if not file.name.lower().endswith('.pdf'):
+        raise gr.Error("Please upload a PDF file.")
+    
+    try:
+        pdf_reader = pypdf.PdfReader(file.name)
+        text = ""
+        for page in pdf_reader.pages:
+            text += page.extract_text()
+    except Exception as e:
+        raise gr.Error(f"Error reading the PDF file: {str(e)}")
+    
+    # Truncate text to 2048 tokens
+    truncated_text = truncate_text(text)
+    if len(truncated_text) < len(text):
+        print("Warning: The input text was truncated to fit within 2048 tokens.")
+    
+    # Generate script
+    try:
+        script = generate_script(SYSTEM_PROMPT, truncated_text, tone)
+    except Exception as e:
+        raise gr.Error(f"Error generating script: {str(e)}")
+    
     # Generate audio for each dialogue item
+    audio_segments = []
+    transcript = ""
+    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 segments
-    # Return audio file and transcript
+    combined_audio = sum(audio_segments)
+    
+    # Adjust length if needed
+    if length == "Short (1-2 min)":
+        combined_audio = combined_audio[:120000]  # 2 minutes max
+    else:  # "Medium (3-5 min)"
+        combined_audio = combined_audio[:300000]  # 5 minutes max
+    
+    # Save combined audio to a temporary file
+    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
 
-# Gradio interface
 instructions = """
 # Podcast Generator
 
 Welcome to the Podcast Generator project! This tool allows you to create custom podcast episodes using AI-generated content.
 
 ## Features
-* Generate podcast scripts on any topic
+* Generate podcast scripts from PDF content
 * Convert text to speech for a natural listening experience
 * Choose the tone of your podcast
 * Export episodes as MP3 files
 
 ## How to Use
-1. Upload a PDF file (max 2048 tokens)
+1. Upload a PDF file (content will be truncated to 2048 tokens if longer)
 2. Select the desired tone (humorous, casual, formal)
 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 gTTS for text-to-speech conversion. The input is limited to 2048 tokens to ensure compatibility with the model.
 """
 
 iface = gr.Interface(
     fn=generate_podcast,
     inputs=[
-        gr.File(label="Upload PDF file"),
+        gr.File(label="Upload PDF file", type="file"),
         gr.Radio(["humorous", "casual", "formal"], label="Select podcast tone", value="casual"),
         gr.Radio(["Short (1-2 min)", "Medium (3-5 min)"], label="Podcast length", value="Medium (3-5 min)")
     ],
@@ -42,8 +92,11 @@ iface = gr.Interface(
         gr.Audio(label="Generated Podcast"),
         gr.Markdown(label="Transcript")
     ],
-    title="Custom NotebookLM-type Podcast Generator",
-    description=instructions
+    title="Custom NotebookLM-type Podcast Generator (2048 token limit)",
+    description=instructions,
+    allow_flagging="never",
+    theme=gr.themes.Soft()
 )
 
-iface.launch()
+if __name__ == "__main__":
+    iface.launch()