ssh-ldn

London Noise Map for renting, buying, and visiting.

Live demo → https://ssh-ldn.app · GitHub

Built in one day at Londonmaxxing 003 (4 July 2026) — Live London track.

The problem

People can estimate commute times, crime rates, and rent — but they have almost no tools for knowing whether a flat will actually sound comfortable to live in. Train vibration, pub gardens, aircraft paths, and weekend nightlife often only show up after move-in day.

ssh-ldn makes invisible sound risks legible: search an address, see what's contributing to noise, and ask follow-up questions by voice.

What it does

• Search any London address or postcode — geocoded via Nominatim / postcodes.io
• Interactive noise map — DEFRA strategic road, rail, and airport layers on a MapLibre base map
• Local sources — pubs, clubs, venues, and hospitals from OpenStreetMap, with day/night/weekend time slots
• Noise score & breakdown — dominant sources, time profile, confidence band, and recommended checks
• Representative sound preview — cursor-point ambient mix (illustrative, not measured audio)
• Voice mode — ask "How noisy is this after 10pm?" via ElevenLabs Speech Engine (accessibility-first)

Try it

Address	What to expect
5/7 Euston Rd., London NW1 2SA	Very noisy — King's Cross / Euston Road
78–80 Wapping Ln, London E1W 2RT	Quieter road/rail; aircraft context
77A Charterhouse St, London EC1M 6HJ	Fabric — strong day vs night pattern
33 Rosaville Rd, London SW6 7BN	Residential Fulham — relatively quiet

Data sources

Source	Use
DEFRA Round 4 strategic noise maps	Road, rail, airport baseline (2021, 10 m grid)
OpenStreetMap / Overpass	Nightlife venues, hospitals, rail / tube / Overground / Elizabeth / DLR / Tram geometry
OpenFreeMap	Vector Positron / Dark Matter tiles
postcodes.io	Postcode autocomplete
TfL Unified API	Station names, zones, and line colours — committed snapshot in data/transit (live keys not required)
Planning London Datahub	Nearby development applications

Strategic noise maps show annual averages, not live measurement. Local opening hours are partial; time-slot activity uses heuristics when hours are missing.

Stack

Next.js 16 · React 19 · MapLibre GL · Tailwind 4 · shadcn/ui · ElevenLabs Speech Engine · Vercel

Local development

pnpm install
cp .env.example .env.local
pnpm dev

Runs on http://localhost:3999.

Live TfL API keys are not required — station points and line identity come from committed data/transit JSON. Add ElevenLabs keys only if you re-enable voice mode (VOICE_MODE_ENABLED and NEXT_PUBLIC_VOICE_MODE_ENABLED). Google geocoding keys are optional. Voice routes in development proxy to the deployed server. See .env.example for the full list.

pnpm typecheck
pnpm lint
pnpm build

Vercel Web Analytics is enabled via @vercel/analytics in app/layout.tsx (no env var required on Vercel).

POI density tiles

Zoomed-out local-source rendering uses prebuilt raster PNG tiles in public/poi-density/tiles/, generated from OSM pubs, bars, nightclubs, music venues, and hospitals:

pnpm generate-poi-density
pnpm generate-poi-density -- --from-cache

The generator writes public/poi-density/manifest.json with the 0–1 normalisation range, per-amenity weights, and tile counts. Live OSM emoji markers still load progressively for higher-zoom detail.

Transit overlays

Tube, Overground, Elizabeth line, DLR, and Tram geometry is snapshotted into data/transit/:

pnpm snapshot-transit
pnpm audit-osm-rail

snapshot-transit refreshes the committed unique-track JSON. audit-osm-rail reports duplicate OSM rail geometries. The map reads these files; it does not call TfL at runtime for line identity.

Accessibility

The map is visual-first, but the product is designed so sound environments can be described, not just shown. Voice mode lets users ask natural questions about a specific address — useful for anyone who navigates the world primarily by hearing, or who wants faster answers than clicking through layers.

Hackathon pitch checklist

From our planning notes — before presenting, make sure these are visible:

• GitHub repository
• Live demo URL — https://ssh-ldn.app
• QR code (repo or demo)
• LinkedIn QR
• Team information
• Data sources (footer on map + this README)
• Architecture (this README + doc/)
• Accessibility story (voice mode)
• Personal motivation — "Everyone remembers moving in and discovering something nobody warned them about."

Licence

Application code: see repository licence. Map data attributions appear in the app footer and follow each provider's terms (OGL v3.0, ODbL, etc.).