app.py

# from utils.subtitles import getSubs

# subs = getSubs("G8gEos8F9R0")
# print(subs)

# import json

# with open("subs.json", "w") as f:
#     json.dump(subs, f)

import datetime
import gradio as gr

def greet(name):
    pass

def run_model(video_id, outname, chunk_size):
    CHUNK_SIZE = chunk_size
    OUT_PPT_NAME = outname
    from rich.progress import track
    from utils.subtitles import getSubsText
    from models.distilbart_cnn_12_6 import summarize
    from models.t5_small_medium_title_generation import t5model as generate_title
    from utils.marp_wrapper import marp
    import utils.markdown as md
    # from utils.chunk import LangChainChunker as chunker
    from utils.subtitles import subs as chunker
    from utils.ppt import generate_ppt
    from utils.video import video
    import os
    
    # check for marp command
    if os.system("command -v marp >> /dev/null"):
        print("Marp not found. Please install marp-cli.")
        os.system("sudo bash setup.sh")
    
    # Intermediary Markdown file
    print("Creating Markdown file...")
    ppt = marp("summary.md")
    ppt.add_header(
        theme="uncover",
        background="",
        _class="invert",
    )
    
    # smaller font size (1.5rem)
    ppt.add_body("<style> section { font-size: 1.5rem; } </style>")
    
    # Generate video
    vid = video(f"https://youtu.be/{video_id}",
                f"out/vid-{video_id}")
    vid.download()
    
    # Get the Subtitles from the YouTube video
    print("Getting subtitles...")
    
    chunker_init    = chunker(video_id)
    chunks          = chunker_init.getSubsList(size=CHUNK_SIZE)
    chunk_len       = len(chunks)

    print(f"subtitles divided to {chunk_len} chunks")

    chunk_num = 1
    for chunk in track(chunks, description="Processing chunks"):
        print(f"processing Chunk: {chunk_num}/{chunk_len}")
        timestamp = str(datetime.timedelta(seconds=chunk[1]))
        # TODO: better file path
        img_path  = f"out/vid-{video_id}_{timestamp}.png"
        
        summary = summarize(chunk[0])
        vid.getframe(timestamp)
        title = generate_title(summary)

        ppt.add_page( md.h2(title), summary )

        if os.path.exists(img_path):
            ppt.add_body(md.image( img_path,
            align="left", setAsBackground=True, size="contain"))
            
        ppt.marp_end()
        chunk_num += 1
        continue

    print(f"Generating {OUT_PPT_NAME}..")
    ppt.close_file()
    generate_ppt("summary.md", OUT_PPT_NAME)
    
    # return full path to the ppt file
    return os.path.abspath(OUT_PPT_NAME)


demo = gr.Interface(fn=run_model, inputs=["text", "text", gr.Slider(200, 1000)], outputs="file")
    
demo.launch()