app.py

import gradio as gr
import tempfile, subprocess, requests
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain.chat_models import ChatOpenAI
from gtts import gTTS
from bs4 import BeautifulSoup
from PIL import Image, ImageDraw

# OpenAI LLM (fast + accurate)
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.3)
summary_prompt = PromptTemplate.from_template("""
Provide a crisp, promotional-style summary (under 50 words) of the following:

{text}

Summary:
""")
summary_chain = LLMChain(llm=llm, prompt=summary_prompt)

# Extract relevant content from article
def extract_main_content(url):
    resp = requests.get(url, timeout=10)
    soup = BeautifulSoup(resp.content, "html.parser")
    for tag in soup(["nav", "header", "footer", "aside", "script", "style", "noscript"]): tag.decompose()
    paras = [p.get_text() for p in soup.find_all("p") if len(p.get_text()) > 60]
    return "\n".join(paras[:20]) or None

# Gradient background
def create_background(image_path, size=(1280,720)):
    img = Image.new("RGB", size)
    draw = ImageDraw.Draw(img)
    for y in range(size[1]):
        color = (10 + y//10, 20 + y//12, 50 + y//15)
        draw.line([(0, y), (size[0], y)], fill=color)
    img.save(image_path)

# AV generation logic with variable duration
def url_to_av_summary(url, duration):
    content = extract_main_content(url)
    if not content:
        return "Failed to extract article content.", None
    summary = summary_chain.run(text=content[:3000]).replace('"','')[:300]

    audio_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name
    gTTS(text=summary).save(audio_path)

    bg_path = tempfile.NamedTemporaryFile(delete=False, suffix=".png").name
    create_background(bg_path)

    video_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4").name
    cmd = [
        'ffmpeg', '-y',
        '-loop', '1', '-i', bg_path, '-i', audio_path,
        '-vf', (
            "drawtext=fontfile=/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf:text='" + summary +
            f"':fontcolor=white:fontsize=48:box=1:[email protected]:boxborderw=5:x=(w-text_w)/2:y=h-(t*(h+text_h)/{duration})"
        ),
        '-t', str(duration),
        '-c:v', 'libx264', '-c:a', 'aac', '-pix_fmt', 'yuv420p', '-shortest', video_path
    ]
    subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)

    return summary, video_path

iface = gr.Interface(
    fn=url_to_av_summary,
    inputs=[
        gr.Textbox(label="Article URL"),
        gr.Radio([5, 10], label="Video Duration (sec)", value=5)
    ],
    outputs=[
        gr.Textbox(label="Summary"),
        gr.Video(label="Generated AV Summary")
    ],
    title="🎞️ AV Summary Generator (OpenAI Powered)",
    description="Generate a short AV video (5 or 10 seconds) summarizing any article. Uses OpenAI + gTTS + FFmpeg."
)

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