CinePilot AI 🎬

Inspiration

Film pre-production is complex and time-consuming. Teams spend weeks breaking down screenplays manually. I wanted to build an AI agent that automates this process and makes production planning accessible to everyone.

What it does

CinePilot AI transforms screenplay text into structured production intelligence:

• Scene detection and character extraction
• Location and time-of-day analysis
• Props, wardrobe, sound, and lighting extraction
• Production budget estimation
• AI casting suggestions
• Script coverage reports

How I built it

Built with Python using Google Gemini agent with tool calling, semantic RAG for screenplay search, and ClickHouse MCP for production data storage. The agent combines deterministic screenplay analysis tools with Gemini's reasoning capabilities.

Challenges

• Separating screenplay facts from production inferences
• Managing Gemini API quotas for embeddings
• Ensuring accurate extraction across different screenplay formats

What I learned

Building an agentic workflow that combines deterministic tools with AI reasoning creates more reliable results than pure AI generation. The key is grounding AI outputs in explicit screenplay data.

Built With

• javascript
• solidity