app.py

import gradio as gr
from moviepy import VideoFileClip, TextClip, CompositeVideoClip
from PIL import Image, ImageDraw, ImageFont
import whisper
from keybert import KeyBERT
import numpy as np
import warnings
import os

# Suppress Whisper FP16 CPU warning
warnings.filterwarnings("ignore", category=UserWarning, module="whisper")

# Load models
whisper_model = whisper.load_model("base")
kw_model = KeyBERT()

# Optional: Hindi-compatible system font (must be pre-installed or accessible by name)
SYSTEM_FONT_NAME = "Noto-Sans-Devanagari"  # This should be installed on the OS

def process_video(video_path, caption="Your Caption"):
    # === 1. Load Video ===
    clip = VideoFileClip(video_path)

    # === 2. Extract Frame at 5s for Thumbnail ===
    frame = clip.get_frame(5)
    image = Image.fromarray(np.uint8(frame))
    draw = ImageDraw.Draw(image)

    # Use PIL to draw caption on thumbnail
    try:
        pil_font = ImageFont.truetype("/usr/share/fonts/truetype/noto/NotoSansDevanagari-Regular.ttf", size=40)
    except:
        pil_font = ImageFont.load_default()

    draw.text((50, image.height - 100), caption, fill="white", font=pil_font)

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

    # === 3. Transcribe Audio in Hindi ===
    result = whisper_model.transcribe(video_path, language="hi")
    text = result["text"]

    # === 4. Extract SEO Keywords ===
    keywords = kw_model.extract_keywords(text, keyphrase_ngram_range=(1, 2), stop_words='english')
    keywords_list = [kw[0] for kw in keywords]

    # === 5. Burn Caption onto Video ===
    try:
        text_clip = TextClip(
            caption,
            fontsize=50,
            color='white',
            font=SYSTEM_FONT_NAME,  # Use font name (not file path)
            method='caption',
            size=(clip.w, None)
        )
    except Exception as e:
        print("Font fallback used due to error:", e)
        text_clip = TextClip(caption, color ="black") 
        

    text_clip = text_clip.set_position(("center", "bottom")).set_duration(clip.duration)
    final_video = CompositeVideoClip([clip, text_clip])

    output_path = "output_with_caption.mp4"
    final_video.write_videofile(output_path, codec="libx264", audio_codec="aac", verbose=False, logger=None)

    return thumbnail_path, ", ".join(keywords_list), output_path

# === Gradio Interface ===
with gr.Blocks() as demo:
    gr.Markdown("# 📹 Video Caption & SEO Tool (Hindi Supported)")

    with gr.Row():
        video_input = gr.File(label="📁 Upload Video", type="filepath")
        caption_input = gr.Textbox(label="📝 Caption for Thumbnail & Video", value="यह शानदार वीडियो है!")

    generate_button = gr.Button("🚀 Generate")

    with gr.Row():
        thumbnail_output = gr.Image(label="🖼️ Generated Thumbnail")
        keywords_output = gr.Textbox(label="🔑 Extracted SEO Keywords")

    video_output = gr.File(label="⬇️ Download Final Video")

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

demo.launch(share=True)