CineForge: Multi-Style Video Caption Generator

CineForge is an AI-powered video caption generation application that transforms short videos into engaging captions across multiple writing styles. Users upload a video through an intuitive Streamlit interface, after which the application automatically extracts representative key frames using FFmpeg. These frames are analyzed by Google's Gemma vision-language model to build a comprehensive scene memory that captures the setting, subjects, actions, visible text, and temporal flow of the video. Instead of generating captions independently for each frame, CineForge first creates a structured understanding of the entire scene. This shared scene memory is then used to produce multiple caption styles, including Formal, Sarcastic, Tech Humor, and Casual Humor, ensuring that every caption remains contextually consistent while matching different audiences and platforms. The application is designed with a modular architecture consisting of separate components for video processing, prompt engineering, multimodal inference, and caption generation. This design makes the pipeline easy to maintain, extend, and adapt to future caption styles or multimodal models. CineForge demonstrates how multimodal AI can bridge computer vision and natural language generation to create content that is both descriptive and creative. It reduces the manual effort required to write engaging captions for social media, marketing content, educational videos, and digital storytelling while showcasing the capabilities of modern vision-language models. The project is containerized using Docker for reproducible deployment and provides a clean, interactive user experience through Streamlit, making advanced multimodal caption generation accessible to both technical and non-technical users.