terminal

Cyril Nwachukwu

Back to Projects
PythonLangGraphGemini 2.5 Flash

PropVid: AI Real Estate Listing Agent

Realtors in Nigeria can go from raw property photos to a published YouTube listing video in under 10 minutes, with zero manual editing.

report_problemThe Problem

Creating high-quality real estate video tours involves tedious scripting, recording voiceovers, and aligning images manually. For agents looking to scale, this process is an expensive bottleneck that takes up to 48 hours per listing. The challenge was building an end-to-end multi-modal pipeline requiring zero human intervention.

psychologyThe Solution

We engineered PropVid using LangGraph's multi-agent architecture. The system orchestrates multiple steps: Gemini 2.5 Flash analyzes the uploaded photos to identify key selling features, a writing agent crafts a targeted sales script, and ElevenLabs generates a natural voiceover. Finally, Creatomate/Cloudinary pipelines render the entire video and publish it via the YouTube Data API.

Project Demo

The Results

Editing Workload

Zero

Manual editing steps required.

Turnaround Time

< 10 min

From photo upload to YouTube.

Process Stability

100%

Automated full pipeline completion.