Skip to main content
v2026.11,986 entries · CC-BY 4.0

Editorial · CASRAI · AI and ML research outputs

University of Richmond Distant Viewing Lab Releases DataReel AI Film Research Tool

The University of Richmond Distant Viewing Lab has released DataReel, an open AI platform analyzing cinematography across 2,420 films and 156 TV series.

Published 21 Nov 2026· 7 minute read
Motion picture film reel in an archival preservation container
Photo: Archives New Zealand from New Zealand / Wikimedia Commons, CC BY 2.0 (cropped). Source · Licence

On November 20, 2026, the University of Richmond’s Distant Viewing Lab publicly released DataReel, an open-access artificial intelligence platform that extracts and visualizes formal cinematic structures across 2,420 films and 156 television series, establishing a standardized computational framework for university research libraries, digital humanities laboratories, and media scholars studying moving images at scale.

Public Release of the DataReel Film Analysis Platform

The Distant Viewing Lab at the University of Richmond announced the general release of its DataReel platform on November 20, 2026, marking a significant milestone in computational media studies. Directed by Dr. Taylor Arnold, professor of data science and statistics, and Dr. Lauren Tilton, professor of digital humanities and director of the Center for Liberal Arts and AI, the system applies multimodal computer vision and natural language processing models to deconstruct moving image collections without distributing copyrighted video files. The initiative is backed by a grant awarded by the Humanities and AI Virtual Institute at Schmidt Sciences, in collaboration with researchers at the University of California, Berkeley, and Bowdoin College.

Traditional film analysis and visual media scholarship have historically relied on manual close viewing, where scholars transcribe camera angles, shot lengths, and character appearances by hand. DataReel automates this workflow by transforming video files into multi-dimensional statistical profiles. The web-based system, accessible at the DataReel interactive visualization portal, provides scatter plots, filtering tools, and visual contact sheets that allow researchers to examine visual style across decades of cinematic production. By publishing derived numerical metrics rather than raw video streams, the project addresses long-standing intellectual property and fair-use bottlenecks that have constrained large-scale digital humanities research across university campuses.

Corpus Scope and Computational Architecture

The public release encompasses automated annotations for 2,420 feature films and 156 television series spanning silent cinema, classical Hollywood, international cinema, and contemporary broadcast television. Each moving image asset is segmented algorithmically into individual shots, which are then analyzed across multiple aesthetic and structural dimensions. Computational models extract camera distance, camera motion, onscreen headcounts, and acoustic proportions, creating a comprehensive registry of research data for computational film studies. Complete tabular datasets are made accessible through the Distant Viewing Lab data repository.

To capture narrative progression alongside visual composition, the engineering architecture incorporates multimodal language models that evaluate dialogue transcripts, intertitles, and plot summaries to divide works into thematic scenes. Face-tracking algorithms follow actors across shot transitions and correlate detected facial features with external photographic cast registries. The derived data points are compiled into compressed, structured tables formatted for immediate ingestion into statistical programming environments such as Polars in Python or tidyverse packages in R.

Analytical Dimension Technical Pipeline Measured Metric or Unit Corpus Coverage
Shot Segmentation Algorithmic boundary detection Duration in seconds; median shot length (MSL) 2,420 films and 156 series
Framing Distance Computer-vision human pose estimation Scale: extreme close-up to extreme long shot Shot-by-shot across all titles
Camera Mobility Optical flow and motion vector analysis Camera motion intensity index Shot-level motion classification
Audio Composition Acoustic signal separation Speech share vs. music share (ratio 0.0 to 1.0) Full runtime audio tracks
Cast Tracking Facial recognition and photo matching Character presence and screen-time duration Matched against credited cast lists
Scene Modeling LLM semantic dialogue segmentation Thematic scene descriptions and keyframe sheets All episodic and feature scenes

Background in Distant Viewing and Cultural Analytics

DataReel represents the practical realization of methodological concepts first introduced in Arnold and Tilton’s 2023 monograph, Distant Viewing: Computational Exploration of Digital Images, published by the MIT Press. Building upon the foundational concept of distant reading pioneered in literary studies by Franco Moretti, distant viewing provides an explicit mathematical and interpretive grammar for converting visual culture into structured evidence. Earlier iterations of the lab’s work produced specialized tools such as the Distant Viewing Toolkit with support from the National Endowment for the Humanities and the Andrew W. Mellon Foundation.

While computer vision object detectors have become standard in industrial computer science, applying these techniques to moving image history presents unique epistemological hurdles. Commercial machine vision models are typically optimized to classify isolated physical objects rather than cinematic grammar—such as editing tempo, shot-countershot dialogue staging, or chiaroscuro lighting. In trade coverage published by Superpower Daily, analysts noted that DataReel directly bridges this divide by turning film into measurable parameters while leaving the interpretive questions of narrative pacing and affective audience response in the hands of humanistic scholars.

Furthermore, DataReel demonstrates how the FAIR data principles—findability, accessibility, interoperability, and reusability—can be systematically implemented within humanities disciplines. By aligning variable definitions with a rigorous codebook and standardizing tabular outputs, the lab provides an empirical baseline for replication and comparative inquiry across disparate archival collections.

Institutional Collaboration and Critical Reception

The creation of DataReel was supported through a $750,000 multi-institution grant from the Schmidt Sciences Humanities and AI Virtual Institute, with $250,000 allocated directly to the University of Richmond. The broader research consortium includes co-principal investigators David Bamman, associate professor of information at UC Berkeley, and Allison Cooper, associate professor of cinema studies at Bowdoin College. This interdisciplinary partnership integrates computer science, statistical modeling, and film history to ensure that computational models reflect historical film theory rather than simplistic pattern matching.

The Distant Viewing Lab hosted preliminary findings at the HAVI Symposium at Bowdoin College in September 2026, presenting pilot analyses of television pacing and historical framing trends. In institutional statements accompanying the launch, Tilton emphasized that contemporary artificial intelligence models have largely excelled at recognizing two-dimensional objects but struggle to understand how or why filmmakers employ specific stylistic choices. Arnold noted that releasing standardized measurement tables allows researchers to test quantitative hypotheses regarding genre evolution, visual pacing, and sound design across vast institutional corpora.

Implications for University Libraries, Humanities Labs, and Research Offices

For research administration leaders, university librarians, and digital scholarship directors, DataReel demonstrates a sustainable model for ethical AI research involving copyrighted cultural heritage materials. Higher education institutions frequently face legal barriers when licensing commercial film and television libraries for text and data mining (TDM). Copyright restrictions often prohibit universities from redistributing digitized video files to external collaborators or publishing complete audiovisual corpora online.

The DataReel methodology circumvents these restrictions through a non-consumptive research model. By generating structured numerical annotations and publishing them under open terms, institutions can share analytical outputs without violating copyright laws. University libraries can adopt this approach as a blueprint for curating specialized collections, leveraging a consistent metadata schema to index audiovisual assets across institutional repositories.

Moreover, the project offers valuable operational lessons for academic research integrity and reproducible computing. Because moving image files consume significant server bandwidth and storage infrastructure, sharing lightweight tabular summaries accompanied by a structured datasheet for datasets dramatically lowers computational overhead. Researchers can replicate complex cinematic inquiries on standard laptops using lightweight data science libraries without needing specialized high-performance computing clusters or enterprise cloud storage subscriptions.

Upcoming Milestones and Methodological Next Steps

Following the public deployment of DataReel, the Distant Viewing Lab has scheduled several training initiatives to help humanities faculty, graduate students, and information professionals incorporate computational video analysis into ongoing scholarship. Detailed program schedules and registration forms are published on the Distant Viewing Lab workshops calendar. The lab will host a virtual Mellon Distant Viewing Workshop from February 15 to February 18, 2027, followed by a free, open online mini-course running February 25 and 26, 2027, dedicated to machine learning methods for visual culture.

An intensive in-person Mellon workshop will take place at the University of Richmond campus from March 8 to March 11, 2027, with additional regional training sessions planned for Fall 2027. On the technical side, the research team plans to expand the corpus to include broader international television broadcast collections and non-Western cinematic traditions, while refining zero-shot visual framing models to track complex shot transitions and lighting designs across historical media archives.

Sources

Ask CASRAI · full access

Ask about this story

Full access: large monthly usage, up to 10 source passages behind each answer, and MCP and API access for Claude Code, Codex and other CLIs. Cancel any time.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →