StudioOS

Film sets can cost ₹50,000–₹5L per hour. Disruptions like rain, illness, or equipment failure force managers to make costly decisions under intense pressure.

Comment

Inspiration

"Stage 3 generator just blew. We're losing ₹1,00,000 every 15 minutes. WHAT DO WE DO?!"

On a live film set, chaos costs a fortune. When equipment breaks, Assistant Directors panic, scribble on scrap paper, and attempt emergency mental math under extreme pressure.

The result? Disaster. They accidentally violate union labor contracts:

  • Skipping mandatory 6-hour crew meal windows (Shooting time > 6 hrs), triggering compounding financial penalties.
  • Blowing the non-negotiable 12-hour turnaround rest (Rest window < 12 hrs) before tomorrow’s 07:00 AM call, triggering forced-call double wages.

We asked: Why are multi-crore film productions still managed like it’s 1995? Can AI rescue a live shooting day in under 3 seconds without hallucinating union rules?

Enter StudioOS.

What it does

StudioOS is an autonomous, multi-agent operational co-pilot for film sets. When an on-set disruption is logged, a 4-agent swarm resolves the crisis in real time:

  • 🎬 Director Agent: Triages incident classification (Type ∈ {Equipment, Weather, Actor, Location}) and launches the specialized agent pipeline.
  • 📅 Scheduling Agent: Reshuffles call sheets, swaps to covered sets, and enforces the hard grace period boundary (Grace Period ≤ 12 min, max once per day).
  • 💰 Budget Agent: Computes the exact rupee cost impact in real time using contractual union formulas (Meal Cost = Crew Count × Tier Rate(Delay)).
  • ⚠️ Risk Agent: Catches cascading downstream traps by checking next-day rest intervals (Rest Window = Tomorrow Call Time - Today Wrap Time ≥ 12 hrs).
  • 📋 Producer Board: Delivers a mathematically verified action plan on a tactile clipboard interface with one-click Approve (elevating confidence to 100%), Reject, or Show Alternative.
  • 🥊 The Benchmark Killer: Ask ChatGPT or a single raw LLM prompt to solve this, and it hallucinates numbers, misses meal penalties, and costs you > ₹21,000 in contract fines. StudioOS pairs LLM reasoning with pure deterministic rule functions for 100% mathematical precision.

How we built it

We engineered StudioOS like precision cinematic hardware:

  • Frontend: Next.js 16 (App Router), React 19, TypeScript 5, and Tailwind CSS styled with a vintage camera viewfinder and physical brass paper clipboard aesthetic.
  • Physics & Motion: Framer Motion powering Skiper 52 hover mechanics — station cards dynamically expand (flex ratio: 1.0 → 3.5, min-width ≥ 290px) with gold viewfinder ticks while gracefully compressing sibling nodes.
  • Backend & Realtime: PostgreSQL on Supabase with live Realtime table streaming (Latency ≈ 0 ms, zero UI polling).
  • AI Brain + Pure Rules: Anthropic Claude 3.5 Sonnet (with Google Gemini 1.5 Pro & Groq Llama 3.3 fallbacks) for high-level semantic reasoning, locked inside a Pure TypeScript Rules Engine (lib/agents/rules.ts) for all financial and labor math.
  • Deterministic Verification: Vitest test suite verifying zero actor double-booking and checking hard budget ceilings (Total Cost ≤ ₹1,00,000).

Challenges we ran into

  • "LLMs are terrible accountants": When LLMs perform union arithmetic, they hallucinate wildly. Our breakthrough was strict rule separation: LLMs write the contextual reasoning; pure TypeScript functions calculate every single rupee and rest minute.
  • UI Expansion Jitter: Making cards expand horizontally on hover without shifting vertical canvas boundaries required pre-allocating an anti-jitter container envelope (min-height = 430px).
  • Race Conditions on Set: Protecting an instant Producer "Approve" click from getting overwritten by asynchronous background LLM streams required dedicated optimistic ref-locking (userDecisionRef).

Accomplishments that we're proud of

  • ₹0 Math Hallucinations: 100% contract-compliant calculations across 50+ crew members and tiered overtime rates (1.5× and 2.0×).
  • Sub-3-Second Crisis Resolution: From generator blowout to signed Producer authorization in ≤ 3 seconds.
  • The Tactile UI Feel: An interface marrying physical Arri camera hardware with an analog production clipboard.
  • The Live Benchmark Engine: A built-in comparative tool proving why single-prompt LLMs fail by > ₹21,000 where StudioOS succeeds.

What we learned

  • High-stakes operations require hard deterministic rails: In workflows governed by union contracts (film, aviation, healthcare), unconstrained prompts are a liability. Pure mathematical functions wrapped in LLM reasoning provide the ideal balance of intelligence and safety.
  • Spatial UI > Chatbots: On a frantic film set, nobody wants a chatbot. Crew members need live telemetry gauges, glowing status nodes, and binary action gates (Approve | Reject).

What's next for StudioOS

  • 📲 Automated WhatsApp Dispatch: Instantly push revised call sheets and scene orders to cast & crew phones via Twilio.
  • 🎙️ Walkie-Talkie AI Listener: Whisper AI monitoring 1st AD radio chatter to detect delays before formal tickets are even typed.
  • 🌐 Multi-Union Rulepacks: Instant toggle between Hollywood (SAG-AFTRA / IATSE), Bollywood (FWICE), and European film union contract rules.

Built With

Share this project:

Updates

Submission history