AI-Powered Screenplay & Storyboard Platform

Context

A web application for film and video pre-production that streamlines the breakdown of screenplays into production elements. The platform uses multiple LLMs (OpenAI, Anthropic, Gemini) to automatically identify and categorize scenes, characters, props, and locations from imported screenplays. Users can then edit and organize storyboards directly in the app.

Role and scope

Full Stack Engineer: added AI integration with multi-LLM provider support, rebuilt the PDF import pipeline, implemented storyboard editing features, and resolved production bugs.

Challenge

Pre-production teams were manually breaking down screenplays, a time-consuming process prone to missed elements. The existing app lacked AI integration, had bugs in the PDF import pipeline, and needed new features to be production-ready.

Approach: Enhanced the platform by integrating multi-LLM support for AI-powered screenplay analysis, fixing and rebuilding the PDF import pipeline, and adding storyboard editing features. The backend runs on FastAPI with async processing and supports switching between LLM providers based on task complexity and cost.

Key decisions and trade-offs

• Added multi-LLM provider support instead of locking to one vendor so cost and quality could be tuned per task.
• Rebuilt the PDF import pipeline rather than patching brittle edge cases, prioritizing reliability over short-term speed.
• Kept AI analysis async so long-running breakdown jobs would not block interactive storyboard editing.

Outcome

• AI breakdown reduces manual categorization from hours to minutes
• Multi-LLM support allows cost-optimized provider selection per task
• PDF import pipeline handles various screenplay formats reliably
• Storyboard editor enables direct visual planning within the platform

Stack

Stack used in this system · Full Stack Engineer