Prompt: ``` Use blender headless to recreate 3D model of this house: and then create a beautiful stunning walkthrough promotion video for the listing. You are an expert cinematographer and animator working in Blender. When creating videos from 3D models, prioritize cinematic quality and visual richness over speed. Be meticulous and thorough. Infer the spatial relationship if needed. The video's camera movement should be bold and fluid, with a cinematic, blockbuster feel. I'd like one-take feel, continuous, unbroken movement. Do visual checks to make sure the 3D model and video look professional quality. Iterate until it is production ready. Target a video length 60 seconds-ish. ``` Disclaimer: I did use @RomeAILab to create the video and it's an open source project ( I built. I already have some rome apps built inside my environment that can pull images from Zillow and add audio to video. The build blocks in the environment helps make the job simpler. For example, I tried claude code with this prompt and it actually failed at the first step because it pulled the wrong photos. It realized the error but still proceeded with the video creation :(

GPT-6 Astra is a beast. Give it a Zillow listing. It can 3D model the house based on the listing photos and create a cool promotion video. The video was just created in one shot and there are still some wrong details. But I am sure it will be much better if I ask it to polish further. OpenAI is so back.

Every app was designed for a million users. Yours should be designed for one. Welcome to Rome OS, an operating system where your agent builds your apps. It builds what it needs to run your life, uses it, sees what works, and improves it. Each pass, your apps fit you a little better. That's the recursive loop. Apps that grow around one person: you. Comment "ROME" and I'll send you early access.