"Don't just predict risks. Predict costs, liabilities, and physics."
ScriptSentinel is an advanced AI-powered "Line Producer" that analyzes film scripts for logistical feasibility, safety risks, and legal liabilities. Unlike standard tools that simply look for keywords (e.g., "Fire" = Danger), ScriptSentinel uses Interpretive Reasoning to understand context, local laws, and physical constraints.
Many scenes look great on paper but are high-risk, expensive, or legally impossible to execute.
- The "Context" Blindspot: Standard AI sees "Rain" and suggests an umbrella. It fails to realize that Rain + Night Shoot + 300 Extras = A logistical nightmare.
- The "Locale" Blindspot: A snow scene is cheap in Kashmir but logistically impossible (and dangerous) in tropical Kerala.
- The "Liability" Blindspot: Most tools ignore local labor laws (e.g., Child Labour Acts) and historical safety precedents.
ScriptSentinel acts as a War Room of three expert AI agents—a Stunt Coordinator, a Lawyer, and a Physicist—debating the script in real-time.
- What it does: Adjusts risk assessments based on the physical realities of the location.
- Example: Flags "Heavy Snow" in Kerala (Tropical) as a Critical Logistics Failure (Heatstroke risk + massive chemical foam cost), whereas in Kashmir, it is marked as Low Risk.
- What it does: Identifies Combinatorial Risks (when two safe things become dangerous together).
- Example: "Rain" is fine. "Night" is fine. "Rain + Night" is a Visibility Hazard requiring specific "Wet-Down" techniques instead of active rain towers.
- What it does: Checks script actions against Local Laws and cites Historical Accidents.
- Example: Detects a "Child Actor" in a "Night Scene" and flags it as a violation of the Child Labour (Prohibition and Regulation) Rules, 2017. Cites the Twilight Zone Movie accident for helicopter risks.
- What it does: Detects "Invisible Risks" like impossible lighting windows.
- Example: A 5-page dialogue scene set at "Sunset" (which lasts only 20 mins) is flagged as a Physics Failure because you will run out of light before shooting is complete.
| Component | Technology Used | Why We Chose It |
|---|---|---|
| Inference Engine | Google Gemini 1.5 Flash | Chosen for its 1M+ token context window (can read full scripts) and superior multilingual reasoning. |
| Frontend / UI | Streamlit | Provides a rapid, reactive "Glassmorphism" interface for zero-latency feedback. |
| Backend Logic | Python 3.10+ | Utilizes strict JSON parsing to force the AI into a deterministic output structure. |
| Deployment | Streamlit Cloud | Serverless architecture for instant scalability and GitHub integration. |
| The "War Room" Dashboard | Multilingual Support (Hindi) |
|---|---|
 |
 |
| Real-time risk badges & mitigation plans | Input Hindi -> Output Hindi analysis |
-
Clone the Repository
git clone [https://github.com/YourUsername/script-sentinel.git](https://github.com/YourUsername/script-sentinel.git) cd script-sentinel -
Install Dependencies
pip install -r requirements.txt
-
Set Up API Key
- Get your FREE API Key from Google AI Studio.
- Open
app.pyand paste your key in theapi_keyvariable (or set viast.secretsfor cloud deployment).
-
Run the App
streamlit run app.py
- Input: User pastes a script scene (in English, Hindi, Malayalam, etc.).
- Context Injection: The user selects a "Shooting Location" (e.g., Hyderabad, Kerala).
- The "War Room" Simulation:
- The Gemini 1.5 Flash model receives a complex System Prompt acting as a Line Producer.
- It cross-references the script against:
- Local Weather/Terrain Physics
- Labor Laws & Safety Guidelines
- Historical Accident Databases
- Structured Output: The AI forces a strict JSON output containing Risk Levels, Mitigation Strategies, and Novelty Scores.
- Rendering: Streamlit parses the JSON and renders the "Glassmorphism" UI with dynamic risk badges.
This project is licensed under the MIT License.
Built with ❤️ by RaknaTech