app.py

import gradio as gr
#from moviepy.editor import VideoFileClip
from moviepy import VideoFileClip
from PIL import Image, ImageDraw, ImageFont
import whisper
from keybert import KeyBERT
import numpy as np

# Load Whisper model and KeyBERT model
whisper_model = whisper.load_model("base")
kw_model = KeyBERT()

def process_video(video_path, caption="Your Caption"):
    # Extract frame at 5 seconds
    clip = VideoFileClip(video_path)
    frame = clip.get_frame(5)  # 5 seconds
    image = Image.fromarray(np.uint8(frame))

    # Add caption
    draw = ImageDraw.Draw(image)
    font = ImageFont.truetype("arial.ttf", 40)
  # Make sure Arial.ttf is available
    text_position = (50, image.height - 100)
    draw.text(text_position, caption, (255, 255, 255), font=font)

    thumbnail_path = "thumbnail.jpg"
    image.save(thumbnail_path)

    # Extract keywords
    result = whisper_model.transcribe(video_path)
    text = result["text"]
    keywords = kw_model.extract_keywords(text, keyphrase_ngram_range=(1, 2), stop_words='english')
    keywords_list = [kw[0] for kw in keywords]

    return thumbnail_path, ", ".join(keywords_list)

# Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("# Video Thumbnail Generator with SEO Keywords")
    video_input = gr.File(label="Upload Video", type="filepath")
    caption_input = gr.Textbox(label="Enter Caption for Thumbnail", value="Awesome Video!")
    generate_button = gr.Button("Generate Thumbnail & Keywords")
    thumbnail_output = gr.Image(label="Generated Thumbnail")
    keywords_output = gr.Textbox(label="SEO Keywords")

    generate_button.click(process_video, inputs=[video_input, caption_input], outputs=[thumbnail_output, keywords_output])

# Launch in Hugging Face Spaces
demo.launch()