Interstellar: Audiovisual Plot Analysis

About the Project

This repository contains a data-driven film analysis of Christopher Nolan's Interstellar.

The project utilizes Python and the marimo notebook framework to correlate low-level audio and video metrics with a defined ground truth of the film's narrative structure (divided into five distinct phases: Earth, Miller's Planet, Mann's Planet, 5th Dimension, and Reunion).

Key Features & Methodology

The analysis is based on three main pillars of automated feature extraction:

• Ground Truth Modeling: A theoretical tension curve mapped via Gaussian peaks to compare computational results against the narrative arc.
• Scene Dynamics: Extraction of cut frequencies and shot lengths using scenedetect to measure pacing.
• Visual Analysis: Frame-by-frame luminance calculation via OpenCV to track lighting and color grading shifts across different planetary settings.
• Audio Analysis: Computation of RMS amplitude and onset strength using librosa to quantify sound design intensity and musical scoring.

Tech Stack

• Framework: marimo
• Audio & Video Processing: librosa, OpenCV (cv2), scenedetect
• Data Handling & Visualization: numpy, pandas, matplotlib, plotly

Project Structure & Data Requirements

To run this notebook, ensure your local directory follows this structure. Due to copyright restrictions, the raw media files are not included in this repository.

├── data/
│   ├── Interstellar.mp4      # Original video file (needs to be provided)
│   └── audio.mp3             # Extracted audio track (needs to be provided)
├── plots/                    # Directory for generated output plots
└── film_plot_analysis.py     # Main marimo notebook