Cine-Engine

• 

  Scene Selection - Approved

• 

  Visual Proof of Concept

• 

  Scene Selection - Rejected

About

CINE-ENGINE is a dual-lens workstation that leverages the unique reasoning and multimodal capabilities of the Gemini 3 series to bridge the gap between creative vision and production reality. At its core, the application utilizes Gemini 3 Pro’s Thinking Mode to simulate a Director’s cognitive process. By allocating a dedicated thinkingBudget, the model doesn't just generate "cool shots"; it reasons through the emotional subtext of a script to ensure that every camera move and lighting choice is narratively motivated.
In parallel, Gemini 3 Flash acts as a pragmatic "Line Producer." It performs high-speed risk audits—scanning for safety, legal, and logistical hazards—to calculate a "Feasibility Score."
This creates a "Marathon Agent" effect where the model uses its long-context memory to detect Continuity Drift, warning the user if their current scene’s ambition outpaces the project’s established production scale.
Finally, Gemini 3 Pro Image synthesizes high-fidelity 16:9 conceptual storyboards. We implement a "Style Lock" mechanism that programmatically injects visual parameters from previous scenes into new image prompts, ensuring a consistent cinematic look across the entire film. This orchestration of Pro, Flash, and Image models creates an autonomous decision-engine that doesn't just help you dream, but helps you shoot.

API & AI Models:

@google/genai SDK
Gemini 3 Pro Preview (Thinking Mode for Directorial Reasoning)
Gemini 3 Flash Preview (Parallel Risk Audit & Logistics Analysis)
Gemini 3 Pro Image Preview (High-Fidelity 1K Cinematic Storyboarding)
Gemini 2.5 Flash Image (Fallback for Standard Mode)
Core Concepts: Parallel Multi-Model Orchestration, Long-Context Memory (Marathon Agents),
Reasoning-first Image Prompting.

Built With

• css3
• html5
• lucide
• react
• tailwind
• typescript-(es6+)