Krishna Cinescribe

CineScribe is a multi-modal video intelligence system that processes video content and produces human-like captions in multiple stylistic voices — formal, sarcastic, humorous with tech references, and humorous for general audiences. It uses a cascade of vision-language models and large language models to understand what's happening in a video and describe it with the right tone. What It Does Given any short video clip — a bustling city intersection, a kitten in a garden, friends at a rooftop table, a nighttime drive through rain — CineScribe: Watches the video by extracting and analyzing key frames Understands the scene, subjects, actions, and temporal progression Summarizes the video into a concise, plain-English narrative Generates four captions, each in a distinct style: Formal — objective and factual, like a documentary narrator Sarcastic — dry, ironic, and lightly mocking Humorous (Tech) — funny through a programming or engineering lens Humorous (Non-Tech) — warm, everyday observational humor