EmlakKlip

Status

Live at emlakklip.com since October 2026

Market

Türkiye — real-estate agents and owners

Our role

Product, engineering and operations, end to end

Stack

TypeScript, React, Mapbox GL, Playwright, ffmpeg, Python, Next.js, Cloudflare

A photo of a land parcel usually shows soil and grass. Buyers cannot see the boundary, the slope or the distance to the road and the sea. Drone footage solves this, but it is slow and expensive for every listing.

• Parcel lookup from the official cadastre by plot number or coordinate, with validation.
• A 2.5D terrain renderer: satellite imagery over a real elevation model, with the parcel drawn from its official geometry.
• A camera planner that keeps the parcel as large as possible inside a safe frame, without overflow.
• Points of interest and distances computed from open map data; AI narration that phrases those figures without inventing new ones.
• Voice-over, word-highlighted subtitles, agent branding, vertical and horizontal formats.
• Ordering, card payments, automatic e-Arşiv invoicing, a render worker queue and an admin panel.

1. 01

   No screen recording

   Every frame is computed from t = frame / fps and captured only when the map, terrain and imagery are ready. Machine speed never changes the video.

2. 02

   Readiness is measured, not assumed

   Map libraries can report "loaded" while terrain data is still missing. Capture waits for terrain separately and rejects frames that fail size and integrity checks.

3. 03

   Checks with negative controls

   Each quality check is tested by forcing the defect it looks for. A check that cannot fail is not trusted.

4. 04

   Provider independence

   Imagery providers sit behind one interface; switching between them is configuration, not code.