Director's Cut

Inspiration

Modern film pre-production involves handling massive amounts of unstructured data—scripts, character sheets, shot lists, and evolving continuity notes. I wanted to build a system where AI crew members can collaborate autonomously without losing track of narrative details.

What it does

Director's Cut is an autonomous script-to-screen pre-production pipeline. Powered by Gemini Enterprise, multi-agent crews break down raw scripts, track real-time continuity errors, and manage production data dynamically at runtime.

How we built it

• AI Core: Gemini models orchestrating multi-agent roles (Script Supervisor, Continuity Checker).
• Long-Term Memory: ClickHouse handling lightning-fast vector and relational queries for real-time state tracking.
• Execution & Hosting: Replit cloud environment for rapid deployment and live orchestration.

Challenges we ran into

Balancing real-time agent memory retrieval with strict narrative continuity constraints across large script datasets.

What we learned

How to effectively leverage ClickHouse MCP servers to give autonomous LLM agents instantaneous access to relational production data.

What's next for Director's Cut

Expanding the agent ecosystem to automatically generate storyboard visual prompts and integrate live text-to-video tools.

Built With

• clickhouse
• fastapi
• google-gemini
• python
• replit