import gradio as gr import tempfile, requests, os, subprocess 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, ImageFont, ImageEnhance import ffmpeg import textwrap import random from urllib.request import urlretrieve UNSPLASH_KEY = "-7tFgMCy_pwrouZrC8mmEIBpyskEyP25e3_Y4vWSvBs" llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.3) summary_prompt = PromptTemplate.from_template(""" Provide a crisp, promotional-style summary (under 100 words) of the following: {text} Summary: """) summary_chain = LLMChain(llm=llm, prompt=summary_prompt) 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 def fetch_unsplash_image(query): url = f"https://api.unsplash.com/photos/random?query={query}&orientation=landscape&client_id={UNSPLASH_KEY}" try: resp = requests.get(url).json() return resp['urls']['regular'] except: return "https://img.freepik.com/free-photo/blue-abstract-gradient-wave-wallpaper_53876-102605.jpg" ASSETS = { "logo": "https://huggingface.co/spaces/csccorner/Link-to-video/resolve/main/csharplogo.png", "graphics": [ "https://img.freepik.com/free-vector/startup-launch-concept-with-rocket_23-2147866180.jpg", "https://img.freepik.com/free-vector/artificial-intelligence-concept-illustration_114360-7307.jpg", "https://img.freepik.com/free-vector/business-goal-achievement-banner_33099-1687.jpg" ] } def download_asset(url): local_path = tempfile.NamedTemporaryFile(delete=False, suffix=".png").name urlretrieve(url, local_path) return local_path def create_slides(text, duration, output_folder, max_lines=6): font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf" font = ImageFont.truetype(font_path, 48) logo_path = download_asset(ASSETS["logo"]) chunks = textwrap.wrap(text, width=36) slides = ["\n".join(chunks[i:i+max_lines]) for i in range(0, len(chunks), max_lines)] per_slide_time = duration / len(slides) slide_paths = [] for i, slide_text in enumerate(slides): query = random.choice(slide_text.split()) bg_path = download_asset(fetch_unsplash_image(query)) graphic_path = download_asset(random.choice(ASSETS["graphics"])) bg = Image.open(bg_path).resize((1280, 720)).convert("RGBA") enhancer = ImageEnhance.Brightness(bg) bg = enhancer.enhance(0.3) draw = ImageDraw.Draw(bg) lines = slide_text.split("\n") lines = [line.encode('utf-8', 'replace').decode('utf-8') for line in lines] total_height = sum([font.getbbox(line)[3] - font.getbbox(line)[1] for line in lines]) + (len(lines)-1)*20 y = max((720 - total_height) // 2, 20) for line in lines: w = font.getbbox(line)[2] - font.getbbox(line)[0] draw.text(((1280 - w) // 2, y), line, font=font, fill="white") y += font.getbbox(line)[3] - font.getbbox(line)[1] + 20 logo = Image.open(logo_path).convert("RGBA") logo = logo.resize((160, int(160 * logo.size[1] / logo.size[0]))) bg.paste(logo, (30, 630 - logo.size[1]), logo) graphic = Image.open(graphic_path).convert("RGBA") graphic = graphic.resize((200, 200)) bg.paste(graphic, (1040, 40), graphic) frame_path = os.path.join(output_folder, f"slide_{i}.png") bg.convert("RGB").save(frame_path) slide_paths.append((frame_path, per_slide_time)) return slide_paths 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.invoke({"text": content[:3000]})["text"].replace('"','')[:300] audio_path = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3").name gTTS(text=summary).save(audio_path) frame_dir = tempfile.mkdtemp() slides = create_slides(summary, duration, frame_dir) concat_txt_path = os.path.join(frame_dir, "slides.txt") with open(concat_txt_path, "w") as f: for path, t in slides: f.write(f"file '{path}'\n") f.write(f"duration {t}\n") f.write(f"file '{slides[-1][0]}'\n") concat_img = os.path.join(frame_dir, "video_input.mp4") subprocess.run([ "ffmpeg", "-y", "-f", "concat", "-safe", "0", "-i", concat_txt_path, "-vsync", "vfr", "-pix_fmt", "yuv420p", concat_img ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) final_video = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4").name video_input = ffmpeg.input(concat_img) audio_input = ffmpeg.input(audio_path) ffmpeg.output(video_input, audio_input, final_video, vcodec='libx264', acodec='aac', pix_fmt='yuv420p', shortest=None ).run(overwrite_output=True, quiet=True) return summary, final_video 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 (Visual Promo Style)", description="Generates a 5/10 sec video summary from article URL with clean typography, visuals, C# Corner logo, and themed illustrations." ) if __name__ == '__main__': iface.launch()