Update app.py

app.py

@@ -1,21 +1,14 @@
 import gradio as gr
+import tempfile, subprocess, requests
 from langchain.chains import LLMChain
 from langchain.prompts import PromptTemplate
-from langchain.llms import HuggingFacePipeline
-from transformers import pipeline
+from langchain.chat_models import ChatOpenAI
 from gtts import gTTS
 from bs4 import BeautifulSoup
-import tempfile
-import requests
-import subprocess
-import concurrent.futures
 from PIL import Image, ImageDraw
 
-# CPU-friendly summarization model
-summary_pipe = pipeline("text2text-generation", model="google/flan-t5-base", device=-1)
-llm = HuggingFacePipeline(pipeline=summary_pipe)
-
-# Prompt for <50-word promotional summary
+# 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:
 
@@ -25,33 +18,29 @@ Summary:
 """)
 summary_chain = LLMChain(llm=llm, prompt=summary_prompt)
 
-# Extract main article content (first 20 meaningful paragraphs)
+# 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()
+    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]
-    content = "\n".join(paras[:20])  # limit to top 20 paragraphs
-    return content or None
+    return "\n".join(paras[:20]) or None
 
-# Create gradient background image
+# Gradient background
 def create_background(image_path, size=(1280,720)):
     img = Image.new("RGB", size)
     draw = ImageDraw.Draw(img)
-    for i in range(size[1]):
-        r = int(10 + (i/size[1])*20)
-        g = int(20 + (i/size[1])*30)
-        b = int(50 + (i/size[1])*50)
-        draw.line([(0, i), (size[0], i)], fill=(r, g, b))
+    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)
 
-# Generate 5s AV summary
-def url_to_av_summary(url):
-    text = extract_main_content(url)
-    if not text:
+# 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
-    text = text[:2000]  # truncate
-    summary = summary_chain.run(text=text).replace('"','')[:250]  # short summary
+    summary = summary_chain.run(text=content[:3000]).replace('"','')[:300]
 
     audio_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name
     gTTS(text=summary).save(audio_path)
@@ -62,35 +51,31 @@ def url_to_av_summary(url):
     video_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4").name
     cmd = [
         'ffmpeg', '-y',
-        '-loop', '1', '-i', bg_path,
-        '-i', audio_path,
+        '-loop', '1', '-i', bg_path, '-i', audio_path,
         '-vf', (
             "drawtext=fontfile=/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf:text='" + summary +
-            "':fontcolor=white:fontsize=48:box=1:[email protected]:boxborderw=5:"
-            "x=(w-text_w)/2:y=h-(t*(h+text_h)/5)"
+            f"':fontcolor=white:fontsize=48:box=1:[email protected]:boxborderw=5:x=(w-text_w)/2:y=h-(t*(h+text_h)/{duration})"
         ),
-        '-t', '5',
+        '-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
 
-# Timeout wrapper: 60s max
-def safe_summary(url, timeout_secs=60):
-    with concurrent.futures.ThreadPoolExecutor() as executor:
-        future = executor.submit(url_to_av_summary, url)
-        try:
-            return future.result(timeout=timeout_secs)
-        except concurrent.futures.TimeoutError:
-            return "⏱️ Processing timed out.", None
+    return summary, video_path
 
 iface = gr.Interface(
-    fn=safe_summary,
-    inputs=gr.Textbox(label="Article URL"),
-    outputs=[gr.Textbox(label="Summary"), gr.Video(label="Video Preview")],
-    title="🎥 5-Second AV Summary (CPU-only)",
-    description="Fast, CPU-only AV summary of a URL. Video capped at 5 seconds to prevent timeouts."
+    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()
+    iface.launch()