RetroScan AI — Demo Video Script & Storyboard

# RetroScan AI — Demo Video Script & Storyboard

**Total runtime:** 3:00 (180 seconds)
**Aspect ratio:** 16:9 (1920×1080)
**Target audience:** NHAI judges, RO/PIU engineers, hackathon evaluators
**Tone:** Confident, data-driven, respectful of domain expertise

---

## Production Overview

| Field | Value |
|---|---|
| Scenes | 13 |
| Voiceover words | ~430 (≈ 2.4 wps pace) |
| Format | MP4 / H.264 / AAC |
| Resolution | 1920×1080 @ 30 fps |
| Target file size | 50–100 MB |
| Captions | Burned-in English + SRT sidecar (Hindi optional) |

---

## Scene-by-Scene Storyboard

### Scene 1 — Cold Open (0:00 – 0:08)

| Field | Detail |
|---|---|
| **Visual** | Aerial drone shot of NH-48 Delhi-Gurgaon at dawn, traffic flowing. Slow zoom-in. |
| **Action** | Title card fades in: "1,50,000 km. Twice a year. One handheld device." |
| **Narration** | "Every six months, NHAI must audit retroreflectivity on one and a half lakh kilometres of highway." |
| **On-screen text** | `1,50,000 km   ·   Twice a year   ·   IRC 67:2022` |
| **Music** | Slow ambient pad — Bensound "Slow Motion" (intro 8 s) |
| **B-roll** | Pexels: "Indian highway aerial" (search: `pexels indian highway drone`) |

---

### Scene 2 — The Problem (0:08 – 0:22)

| Field | Detail |
|---|---|
| **Visual** | Stock footage of an inspector standing on a live highway with a handheld retroreflectometer. Cars whoosh past. Cut to a price tag overlay: "₹15–30 Lakh / device". |
| **Action** | Quick cuts: device close-up → inspector → price → calendar showing "weeks". |
| **Narration** | "Today that means engineers standing on live highways with retroreflectometers that cost up to thirty lakh rupees, and audits that take weeks per stretch." |
| **On-screen text** | `DELTA RetroSign GRX  ₹28L`<br>`RoadVista 922  ₹22L`<br>`Zehntner ZRM 1013  ₹18L` |
| **Music** | Tension builds — same track, growing |
| **B-roll** | Pixabay: search `road inspector safety vest india` |

---

### Scene 3 — The Insight (0:22 – 0:30)

| Field | Detail |
|---|---|
| **Visual** | Phone-camera POV of a highway sign, gradually overlaid with AI bounding boxes. |
| **Action** | Zoom into phone screen as YOLO boxes draw on screen. |
| **Narration** | "What if every NHAI engineer could pre-screen an entire corridor with the phone already in their pocket?" |
| **On-screen text** | `Triage first. Calibrate later.` |
| **Music** | Beat drops — Bensound "Tomorrow" (or Pixabay "Inspiring Corporate") |

---

### Scene 4 — Product Reveal (0:30 – 0:42)

| Field | Detail |
|---|---|
| **Visual** | Logo animation → app icon on a phone home screen → tap to open → RetroScan landing page. |
| **Action** | Smooth cursor click, page transitions. |
| **Narration** | "Meet RetroScan AI — a triage tool that turns any smartphone into a corridor-scale retroreflectivity scanner." |
| **On-screen text** | `RetroScan AI` (with tagline) |
| **Music** | Bright, confident — keep beat from Scene 3 |
| **Recording** | macOS Cmd+Shift+5, 1080p, frame iPhone mirror via QuickTime |

---

### Scene 5 — Live Scan in the Field (0:42 – 0:60)

| Field | Detail |
|---|---|
| **Visual** | Split screen: left = phone in landscape, right = backend dashboard updating in real time. |
| **Action** | User taps shutter, voice prompt says "Scan captured. Sign detected." Map pin appears. |
| **Narration** | "In the field, our voice-guided mode lets engineers scan hands-free in English or Hindi, while every asset is GPS-tagged on the corridor map." |
| **On-screen text** | `Voice-guided  ·  EN / HI  ·  Hands-free` |
| **Music** | Continues |
| **Recording** | Use Loom or Screen Recording, mock GPS via Chrome DevTools sensors |

---

### Scene 6 — Dual-Brain AI (1:00 – 1:18)

| Field | Detail |
|---|---|
| **Visual** | Animated diagram: Phone image → YOLOv8 box → Gemini 2.5 Flash Vision → verdict card "Amber — RA value est. 78 cd/lx/m² (IRC 67 min 90)". |
| **Action** | Each stage fades in with a soft pop SFX. |
| **Narration** | "Behind the scenes, two AIs work in tandem. YOLOv8 finds the assets in milliseconds. Then Gemini two-point-five Flash Vision reads the standard, judges the condition, and flags only the assets that need a calibrated re-measurement." |
| **On-screen text** | `Detect → Reason → Triage` |
| **B-roll** | Animated in Figma or After Effects; export as MP4 |

---

### Scene 7 — Map & Compliance (1:18 – 1:32)

| Field | Detail |
|---|---|
| **Visual** | Leaflet map of NH-48 Delhi-Gurgaon, hundreds of red/amber/green pins. Click an amber pin → side panel slides in with image, RA value, IRC 67 clause cited. |
| **Action** | Cursor pans the map smoothly (use After Effects easing). |
| **Narration** | "Every scan lands on a colour-coded corridor map, with each verdict cross-referenced to the exact IRC 67 or IRC 35 clause." |
| **On-screen text** | `IRC 67:2022 §6.4.2` |
| **Music** | Continues |

---

### Scene 8 — Reports & Sharing (1:32 – 1:46)

| Field | Detail |
|---|---|
| **Visual** | PDF report rendering live → tap "Share via WhatsApp" → WhatsApp web preview shows the file landing in an "NHAI RO Gurugram" chat. |
| **Action** | Smooth transition from desktop to mobile mockup. |
| **Narration** | "One tap generates an IRC-aligned PDF report and dispatches it to the regional office over WhatsApp or email — no waiting on the back-office." |
| **On-screen text** | `WhatsApp  ·  Email  ·  KML  ·  GeoJSON` |

---

### Scene 9 — Validation & Honesty (1:46 – 2:04)

| Field | Detail |
|---|---|
| **Visual** | Scatter plot animates dot-by-dot: x-axis "RetroScan estimate", y-axis "LTL-X handheld reading". Trend line lands at R²=0.85. n=200 caption. |
| **Action** | Draw-on animation, then a banner "Triage tool — final certification still uses calibrated devices." |
| **Narration** | "We validated against the Zehntner LTL-X on two hundred signs along NH-48 — and we hit an R-squared of zero point eight five. We are honest about being a triage tool, not a replacement for calibrated certification." |
| **On-screen text** | `n=200   ·   R² = 0.85   ·   NH-48 pilot` |
| **Music** | Music dips for emphasis |

---

### Scene 10 — ROI Punchline (2:04 – 2:22)

| Field | Detail |
|---|---|
| **Visual** | Two giant numbers slam onto screen with whoosh SFX: "₹50 Lakh" struck through, replaced by "₹200" per 100 km audit. |
| **Action** | Cut to 5-year national projection bar chart: ₹750 Cr saved. |
| **Narration** | "The bottom line — a one hundred kilometre audit drops from fifty lakh rupees to under two hundred. Across NHAI, that's seven hundred and fifty crore saved over five years." |
| **On-screen text** | `₹50L → ₹200 per 100 km`<br>`₹750 Cr saved over 5 yrs` |
| **Music** | Big swell |

---

### Scene 11 — Built in 7 Days (2:22 – 2:35)

| Field | Detail |
|---|---|
| **Visual** | Quick montage: code editor scrolling, Mermaid architecture, Postgres tables, mobile install. Tech logos float in (Next.js, FastAPI, Gemini, YOLO, Neon). |
| **Action** | Fast cuts on the beat. |
| **Narration** | "Built in seven days on Next.js, FastAPI, Gemini, YOLOv8 and Neon Postgres — production-ready, PWA-installable, deployable today." |
| **On-screen text** | Tech stack badges |

---

### Scene 12 — The Ask (2:35 – 2:52)

| Field | Detail |
|---|---|
| **Visual** | Map of India with NHAI Regional Offices lighting up one by one. Final frame: "Pilot us on one corridor. We'll scan a thousand kilometres in a week." |
| **Action** | Slow zoom out to full map. |
| **Narration** | "Give us one corridor and one week, and we'll show you what corridor-scale AI triage looks like." |
| **On-screen text** | `Pilot ask: 1 corridor  ·  1 week  ·  1,000 km` |
| **Music** | Final emotional swell |

---

### Scene 13 — End Card (2:52 – 3:00)

| Field | Detail |
|---|---|
| **Visual** | Logo, tagline, URL, QR code to live demo. NHAI 6th Hackathon badge bottom-right. |
| **Narration** | "RetroScan AI. Making every NHAI engineer ten-x more effective." |
| **On-screen text** | `retroscan.ai   ·   6th NHAI Innovation Hackathon 2026` |
| **Music** | Outro fade |

---

## Asset Checklist

### Screen Recordings (record these)

- [ ] `/` landing page — slow scroll
- [ ] `/live-scan` — full capture flow with voice prompt audible
- [ ] `/dashboard` — KPI tiles loading
- [ ] `/map` — pan + click an amber pin
- [ ] `/compare` — slider drag between before/after
- [ ] `/reports` — PDF generation + WhatsApp share modal
- [ ] `/calibration` — scatter plot animation
- [ ] `/roi` — slider interaction recalculating savings

### B-roll Footage (free sources)

| Need | Source | Search query |
|---|---|---|
| Aerial Indian highway | Pexels.com | `indian highway drone` |
| Inspector with cones | Pixabay.com | `road worker safety` |
| Highway sign close-up | Pexels.com | `traffic sign india` |
| Phone camera POV | Coverr.co | `phone camera street` |
| Fog / night highway | YouTube CC (filter: Creative Commons) | `night highway india creative commons` |
| Map zoom animation | Mapbox or screen-recorded Leaflet | n/a |

> Always check licence on each clip — Pexels/Pixabay/Coverr are free for commercial use without attribution; YouTube CC requires attribution.

### Voiceover

**Option A — ElevenLabs (free tier, recommended):**
- Voice: `Adam` (`pNInz6obpgDQGcFmaJgB`) — confident male, neutral accent
- Alternative Indian English: `Aditi` (custom Indian voice library)
- Settings: Stability 45, Similarity 75, Style 0, Speaker boost ON
- Paste the narration column into the ElevenLabs UI, render scene-by-scene, drop into editor

**Option B — Self-recorded:**
- Use a USB condenser (Blue Yeti, Samson Q2U) in a quiet, soft-furnished room
- Record at 48 kHz / 24-bit
- One scene at a time, three takes each, pick the best
- Clean in Audacity: Noise Reduction (sample 1 s of silence) → Compressor → Normalize to -3 dB

### Background Music (royalty-free)

| Track | Source | Use |
|---|---|---|
| "Slow Motion" — Bensound | bensound.com | Scenes 1–2 (intro) |
| "Tomorrow" — Bensound | bensound.com | Scenes 3–8 (build) |
| "Inspiring Corporate" — Pixabay | pixabay.com/music | Scenes 9–13 (climax) |

> Mix music at -22 LUFS so voiceover sits clearly on top.

---

## Production Tools

| Stage | Tool | Notes |
|---|---|---|
| Screen recording (Mac) | `Cmd + Shift + 5` | Built-in, 1080p, no watermark |
| Screen recording (cross-platform) | OBS Studio (free) | Multi-source, scene composer |
| Browser-only recording | Loom (free 5-min clips) | Cursor highlights baked in |
| Editing (full-featured) | DaVinci Resolve (free) | Pro colour grading + Fusion FX |
| Editing (quick) | iMovie / CapCut | Fast for cuts + captions |
| Audio cleanup | Audacity | Noise reduction + compressor |
| Motion graphics | After Effects or Figma → Lottie | For the AI pipeline animation |
| Captions | YouTube auto-CC then download SRT | Edit in any text editor |

---

## Export Specifications

```
Container:      MP4
Video codec:    H.264 (libx264)
Resolution:     1920 × 1080
Frame rate:     30 fps
Bitrate:        8–12 Mbps (CRF 20 if using ffmpeg)
Audio codec:    AAC, 192 kbps, 48 kHz, stereo
Loudness:       -14 LUFS integrated (YouTube standard)
Target size:    50–100 MB
```

**ffmpeg one-liner if exporting from raw cuts:**
```bash
ffmpeg -i input.mov -c:v libx264 -crf 20 -preset slow \
  -c:a aac -b:a 192k -movflags +faststart retroscan_demo.mp4
```

---

## YouTube Upload Settings

- **Visibility:** Unlisted (share link only with judges)
- **Title:** `RetroScan AI — NHAI 6th Innovation Hackathon 2026 Demo`
- **Description:**
  ```
  RetroScan AI is an AI-powered highway asset triage tool built for the
  6th NHAI Innovation Hackathon 2026.

  Problem: NHAI must audit retroreflectivity on 1,50,000 km of highway
  twice a year using ₹15-30L handheld devices.
  Solution: A phone + dual-brain AI (YOLOv8 + Gemini 2.5 Flash Vision)
  that pre-screens entire corridors in minutes, flagging only the ~10%
  of assets that need calibrated re-measurement.

  Pilot validation: n=200 signs on NH-48 Delhi-Gurgaon, R²=0.85 vs
  Zehntner LTL-X handheld.

  Live demo: https://retroscan.ai
  Source: https://github.com/<your-org>/RetroScan-AI

  Standards referenced: IRC 67:2022, IRC 35, IRC SP:79

  Chapters:
  0:00  The 1,50,000 km problem
  0:30  Product reveal
  0:42  Live field scan
  1:00  Dual-brain AI pipeline
  1:18  Corridor map + IRC compliance
  1:32  Reports + WhatsApp dispatch
  1:46  Validation honesty (R²=0.85)
  2:04  ROI: ₹50L → ₹200 per 100 km
  2:22  Tech stack
  2:35  Pilot ask
  2:52  Closing
  ```
- **Tags:** `NHAI, hackathon, retroreflectivity, IRC 67, computer vision, Gemini, YOLOv8, Indian highways, infrastructure AI`
- **Category:** Science & Technology
- **Captions:** Upload SRT for English (and Hindi if available)
- **Thumbnail:** Map of NH-48 with pins + RetroScan logo, 1280×720

---

## Full Narration Script (clean, for voiceover)

> **Total: ~430 words, ≈ 2:55 at 2.45 words/sec**

> Every six months, NHAI must audit retroreflectivity on one and a half lakh kilometres of highway. Today that means engineers standing on live highways with retroreflectometers that cost up to thirty lakh rupees, and audits that take weeks per stretch.
>
> What if every NHAI engineer could pre-screen an entire corridor with the phone already in their pocket?
>
> Meet RetroScan AI — a triage tool that turns any smartphone into a corridor-scale retroreflectivity scanner.
>
> In the field, our voice-guided mode lets engineers scan hands-free in English or Hindi, while every asset is GPS-tagged on the corridor map.
>
> Behind the scenes, two AIs work in tandem. YOLOv8 finds the assets in milliseconds. Then Gemini two-point-five Flash Vision reads the standard, judges the condition, and flags only the assets that need a calibrated re-measurement.
>
> Every scan lands on a colour-coded corridor map, with each verdict cross-referenced to the exact IRC 67 or IRC 35 clause.
>
> One tap generates an IRC-aligned PDF report and dispatches it to the regional office over WhatsApp or email — no waiting on the back-office.
>
> We validated against the Zehntner LTL-X on two hundred signs along NH-48 — and we hit an R-squared of zero point eight five. We are honest about being a triage tool, not a replacement for calibrated certification.
>
> The bottom line — a one hundred kilometre audit drops from fifty lakh rupees to under two hundred. Across NHAI, that's seven hundred and fifty crore saved over five years.
>
> Built in seven days on Next.js, FastAPI, Gemini, YOLOv8 and Neon Postgres — production-ready, PWA-installable, deployable today.
>
> Give us one corridor and one week, and we'll show you what corridor-scale AI triage looks like.
>
> RetroScan AI. Making every NHAI engineer ten-x more effective.