app.py

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

def generate_podcast(file, url, tone, length):
    if file and url:
        raise gr.Error("Please provide either a PDF file or a URL, not both.")
    
    if file:
        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)}")
    elif url:
        try:
            text = extract_text_from_url(url)
        except Exception as e:
            raise gr.Error(f"Error extracting text from URL: {str(e)}")
    else:
        raise gr.Error("Please provide either a PDF file or a URL.")
    
    truncated_text = truncate_text(text)
    if len(truncated_text) < len(text):
        print("Warning: The input text was truncated to fit within 2048 tokens.")
    
    try:
        script = generate_script(SYSTEM_PROMPT, truncated_text, tone, length)
    except Exception as e:
        raise gr.Error(f"Error generating script: {str(e)}")
    
    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)}")
    
    combined_audio = sum(audio_segments)
    
    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

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 from PDF content or web pages
* 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 OR enter a URL (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 (optional)", file_types=[".pdf"]),
        gr.Textbox(label="OR Enter URL"),
        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)")
    ],
    outputs=[
        gr.Audio(label="Generated Podcast"),
        gr.Markdown(label="Transcript")
    ],
    title="Custom NotebookLM-type Podcast Generator (2048 token limit)",
    description=instructions,
    allow_flagging="never",
    theme=gr.themes.Soft()
)

if __name__ == "__main__":
    iface.launch()