app/utils/whisper_llm.py

import whisper
from transformers import pipeline
import requests
import tempfile


def analyze(video_url: str):
    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
    with requests.get(video_url, stream=True) as r:
        for chunk in r.iter_content(8192):
            tmp.write(chunk)
    tmp.close()

    model = whisper.load_model("base")
    result = model.transcribe(tmp.name)
    text = result["text"]

    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
    summary = summarizer(text, max_length=512, min_length=128, do_sample=False)[0][
        "summary_text"
    ]
    return text, summary