AI Property Tour Video Generator
Turns listing photos into a branded walkthrough video narrated in the agent's own cloned voice.
Overview
Real estate agents wanted listing videos with themselves on screen and their own voice, without paying a videographer for every property. We rebuilt a frontend-only prototype into a full product: upload photos and details, and the system writes the script, generates room-by-room clips, clones the agent's voice and assembles the tour.
The Challenge
Professional listing videos cost $200+ per property and take days to turn around.
The Solution
Simple input
Listing photos, agent photo, property details and a chosen language.
AI script
Written scene by scene to match each room.
Agent compositing and room-by-room clips
Generated from still photos.
Cloned-voice narration
Synced per scene, then the finished tour is packaged for social media.
How It Works
Photos + details
AI script per scene
Agent compositing (Flux)
Room clips (Kling v3)
Voice clone (ElevenLabs)
Assembly → social
Architecture
App
Next.js and TypeScript, with API routes for each generation step.
Script engine
LLM via OpenRouter writes narration per scene.
Image layer
Flux on Fal.ai composites the agent into the scene.
Video layer
Kling v3 on Fal.ai turns stills into room clips.
Voice
ElevenLabs voice clone from a short sample of the agent.
Assembly
Clips and narration stitched into one tour, ready to post.
Engineering Highlights
- Rebuilt from a frontend-only prototype into a product with a real backend pipeline.
- Scene-by-scene generation fixed narration drifting out of sync with the video.
- Clear API error handling so credit or quota issues surface instead of failing silently.
Outcome
- Live with realtors on a monthly subscription.
- Full property tour for roughly $5 to $10 in API cost, versus $200+ for a videographer.