app.py

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

# Initialize OpenAI LLM
llm = ChatOpenAI(model_name="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 top article paragraphs
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(path, size=(1280,720)):
    img = Image.new("RGB", size)
    draw = ImageDraw.Draw(img)
    for y in range(size[1]):
        draw.line([(0,y),(size[0],y)], fill=(10+y//10,20+y//12,50+y//15))
    img.save(path)

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

    # Wrap long summary to multiline
    wrapped_summary = textwrap.fill(summary, width=50).replace("\n", "\\n")

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

    # Background image
    bg = tempfile.NamedTemporaryFile(delete=False, suffix=".png").name
    create_background(bg)

    # Build video stream
    video_stream = ffmpeg.input(bg, loop=1, framerate=1)
    text_opts = dict(
        fontfile="/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf",
        text=wrapped_summary,
        fontcolor="white",
        fontsize=48,
        box=1,
        boxcolor="[email protected]",
        boxborderw=5,
        x="(w-text_w)/2",
        y=f"h-(t*(h+text_h)/{duration})"
    )
    video = video_stream.drawtext(**text_opts).setpts('PTS')
    audio_stream = ffmpeg.input(audio)

    out_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4").name
    out = ffmpeg.output(video, audio_stream, out_path,
                         vcodec="libx264", acodec="aac", pix_fmt="yuv420p", t=duration)
    out.run(quiet=True)

    return summary, out_path

iface = gr.Interface(
    fn=url_to_av_summary,
    inputs=[gr.Textbox(label="Article URL"), gr.Radio([5,10], label="Duration (sec)", value=5)],
    outputs=[gr.Textbox(label="Summary"), gr.Video(label="AV Summary")],
    title="AV Summary Generator",
    description="Generate a promo-style AV summary (5 or 10s) using OpenAI + gTTS + ffmpeg-python."
)

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