Film School Cuts Analysis Time with Agentic AI & Serverless

Executive Summary

A leading European film education institution deployed a fully serverless AWS solution to automate the extraction of film credits from video files. The solution reduced processing time from 30 minutes to 1 minute per film, achieving a 96% improvement and enabling 5x overall efficiency gains.

Customer Background

This prominent film school educates students in film production, directing, cinematography, and post-production while maintaining a national database of film productions and contributors. The institution employs hundreds of students and numerous faculty members involved in both educational and professional film projects, playing a central role in the national film industry by combining education, culture, and technology.

Challenge

The institution needed to automate the extraction of film credits from video files to support its national database of film productions and contributors. Previously, five staff members manually reviewed each film, transcribing end credits line by line; a repetitive, time-consuming process prone to human error and inconsistencies. Without modernization, the institution faced increasing operational costs, slower database development, and logistical bottlenecks during staff absences or resource shortages.

Solution

Trek2Summit built a fully serverless AWS architecture that automatically processes uploaded film files and extracts text from end credits. The workflow is triggered upon file upload and terminates once the output text file is generated:

Amazon S3 - File storage and workflow trigger
AWS Step Functions - Process orchestration with loop iterations for frame extraction and text processing
AWS MediaConvert - Video-to-frame conversion, splitting uploaded videos into individual frames
Amazon Textract - Text detection and extraction from credit frame images
AWS Lambda - Serverless aggregation, duplicate removal, text cleaning, and final output generation
Amazon Bedrock - GenAI-powered generation of film credit summaries and production analyses from extracted data

The solution runs in the Frankfurt region with cost optimization through automatic video file deletion after processing, retaining only the extracted text output. The fully serverless architecture ensures cost-efficiency, scalability, and minimal operational overhead.

Results

96% reduction in processing time (from 30 minutes to 1 minute per film)
5x overall efficiency increase, enabling the team to process significantly more films
100 films processed monthly with capacity for substantial scale-up
Automated workflow removes the majority of manual workload while maintaining quality through manual validation

AWS Services Used