
Automated AI property video production
An end-to-end walkthrough showing how a reusable AI workflow converts a single property listing URL into a finished vertical property advertisement. The system scrapes listing data and images, classifies rooms, auto-generates a scene-by-scene transcript and prompts, selects a consistent virtual agent, synthesizes each clip with generative models, and composes the final video through an orchestrated Remotion render.
– Workflow components: property scraper, image analyzer, transcript generator, character manager, prompt library, generation engine, and Remotion renderer; each implemented as reusable skill files.
– Scene generation: images are organized into a natural tour order, a transcript.json defines scene durations and spoken lines, and prompts combine image, agent sheet, and voice to produce clips via models like C-Dance and High Low.
– Production control: an orchestrator skill validates outputs stage-by-stage, pauses for agent selection, and can retry failed scene generations without redoing completed clips.
– Reuse & delivery: designed to run repeatedly for different listings, exporting a vertical MP4 and saving all project artifacts for review or client delivery.
Quotes:
I’m not real, but this property is.
We only need to build this workflow once — then it can create videos for any listing.
If one scene fails, Codex can retry it without regenerating the complete scenes.
Statistics
| Upload date: | 2026-07-31 |
|---|---|
| Likes: | 52 |
| Comments: | 4 |
| Statistics updated: | 2026-08-02 |
Specification: One Property Listing URL → Full Video With a Virtual Real Estate Agent
|