🎬 INFRAWATCH — Demo Video Script

Target length: 3 minutes (some hackathons cap at 2 — script also has a 2-min cut noted) Format: 1920×1080, screen recording with voiceover (use OBS Studio — free) Tone: Confident, urgent, factual. Not salesy.

---

Pre-recording checklist

• Backend + frontend running locally
• Logged in as admin (admin@infrawatch.gov.in)
• Browser zoom at 100%, dark mode, no other tabs
• Disable browser extensions / notifications
• Pre-position scenes: dashboard loaded, violations list pre-filtered
• Mic test — speak normally, 6 inches from mic
• Have water nearby

---

🎙️ Full Script (3-minute version)

Scene 1 · Opening hook (0:00 – 0:18) — 18 sec

Visual: Black screen → fades to news headlines montage (use a slide with 3 real headlines)

"Dharwad, 2019: 19 people killed when an illegal building collapsed. Bellandur Lake, 2024: still on fire from sewage from unauthorized construction. Bengaluru today: 2.5 million buildings, 198 wards, 840 inspectors. The math doesn't work. India's cities are facing a silent urban safety crisis."

On-screen text overlay:

• "19 lives lost — Dharwad collapse"
• "Lake on fire — Bellandur"
• "840 officers · 198 wards · 2.5M buildings"

---

Scene 2 · Solution intro (0:18 – 0:35) — 17 sec

Visual: Login page fades in, then dashboard loads showing the pulsing red CRITICAL banner

"We built INFRAWATCH — an AI-powered Urban Safety Crisis Response Platform. It detects illegal construction from satellite imagery, files cases automatically, and tracks response times in hours instead of weeks."

On-screen text: "INFRAWATCH · Rapid Crisis Response · Open Innovation"

---

Scene 3 · The Crisis Banner (0:35 – 0:55) — 20 sec

Visual: Zoom into Crisis Response Banner. Show the auto-rotating ticker. Hover/highlight the metric pills.

"The dashboard opens with a live Crisis Banner — right now showing 23 critical cases requiring immediate dispatch. Each case has its own response SLA — 4 hours for high-confidence detections, 12 hours for medium. Watch the ticker auto-rotate through cases breaching their SLAs in real time."

Action: Click on a ticker case to navigate to its detail page (briefly — don't dwell).

---

Scene 4 · The AI City Scan — THE WOW (0:55 – 1:35) — 40 sec

Visual: Back to dashboard. Click the orange "⚡ Run AI City Scan" button. Modal opens.

"Here's what makes this rapid — the AI City Scan."

Action: Click "START AI SCAN" — let the live walkthrough play (~25 sec).

"In under 30 seconds, the system picks four random Bengaluru hotspots, runs live computer vision analysis on satellite imagery using Groq's Llama 4 Scout Vision model, and for every detected violation it auto-creates a case in the database AND auto-assigns the nearest field officer through round-robin dispatch. No human triage. No clipboard. No 21-day waiting period."

Visual: Final summary card with stats — pause for 2 seconds to let viewer read.

---

Scene 5 · End-to-end traceability (1:35 – 2:10) — 35 sec

Visual: Click "View All Violations →" in the modal. Show the violations list with new entries at top.

"Every case the AI detects flows through the complete enforcement pipeline. Here's a violation that was just filed."

Action: Click 🛰 icon on a recent violation. Wait for the Live Detection map to load with the pulsing red pin and the case details overlay.

"Click any case and we jump to the Live Detection map — pin highlighted at the exact GPS coordinates, full case context overlaid, ready for the officer to verify against current satellite imagery."

Action: Click back to violations, then click "View →" on the same case to open detail.

"On the case detail, every violation has a Rapid Response Status card — elapsed time, SLA target, breach status, assigned officer. Plus AI-generated legal notice drafting under BBMP Act Section 321."

Visual: Briefly show the Rapid Response Status card with the colored SLA bar.

---

Scene 6 · Real-world impact (2:10 – 2:35) — 25 sec

Visual: Scroll up to show the Impact Card on the dashboard.

"What does this mean at scale? Today, BBMP processes about 840 violations per year manually. If INFRAWATCH is deployed across all 198 wards, projections show 12,400 violations detected per year — fourteen times more — with detection-to-notice time dropping from 21 days to 2.3 hours. That's 219 times faster. That's the difference between catching a fourth-floor extension while it's a foundation, and arriving after the building has collapsed."

Visual: Highlight the four impact stats one by one (zoom or pulse).

---

Scene 7 · Tech + close (2:35 – 3:00) — 25 sec

Visual: Quick montage — show AI Features chatbot answering a question, then return to dashboard.

"INFRAWATCH is built on React, Express, Groq Vision AI, and is fully open-source under MIT license — ready to be forked for any Indian city. The codebase is at github-dot-com slash YOUR-USERNAME slash infrawatch. A live demo is at infrawatch-dot-app. This is a working prototype. Phase 2 will integrate Google Earth Engine, BBMP's permit database, citizen reporting, and WhatsApp dispatch. Built by students. Engineered for crisis. Open for cities."

Final on-screen text:

• "INFRAWATCH"
• "Open source · MIT"
• "Built for Google Solution Challenge 2026"
• "Theme: Rapid Crisis Response · Open Innovation"

---

✂️ 2-minute cut (if hackathon caps at 2 min)

Drop:

• Scene 5 (end-to-end traceability) — keep only the pin+overlay shot, drop detail page
• Scene 7 — compress to 10 sec, just the close

Keep these in full: Scenes 1, 2, 3, 4, 6.

---

🎤 Voiceover tips

• Don't read like a teleprompter. Practice 3 times so it sounds conversational.
• Pause after big numbers. "Twelve thousand four hundred — (beat) — violations per year."
• Lower your voice on impact lines. Whisper-style works for "the building has collapsed."
• Energy spike on the AI City Scan — that's your wow moment.

---

🎬 Recording setup

Software: OBS Studio (free) — record at 1920×1080, 30fps, MP4 output.

Mic: Use a wired headset or the built-in mic with Krisp (free noise cancellation). Avoid AirPods — they sound tinny.

Screen: Browser only, fullscreen (F11). Hide the bookmark bar.

Cursor: Enable "highlight cursor" in OBS so viewers can follow your clicks.

Editing: DaVinci Resolve (free) or CapCut (free) — just trim, no fancy transitions. Add a 2-second intro card and 2-second outro card.

---

📤 Upload checklist

• YouTube Unlisted (don't make it Public until submission day)
• Title: "INFRAWATCH — Urban Safety Crisis Response | Google Solution Challenge 2026"
• Description: paste your README impact summary
• Thumbnail: dashboard with the red Crisis Banner pulsing
• Add the YouTube link to your hackathon submission

---

⚠️ Common mistakes to avoid

❌ Showing the loading spinner for too long — pre-load everything ❌ Talking over a silent UI — narrate WHILE clicking ❌ "Um", "uh", "so basically" — re-record those takes ❌ Reading the UI text out loud verbatim — the viewer can read ❌ Music louder than your voice — keep music at 15% if used ❌ Going over time limit — judges stop watching at the cutoff

---

Good luck. You've got 3 minutes to convince a judge to remember you out of 1700 teams. Make every second count.