A new US startup aims to take the time and cost out of staging a listing for advertising, instead furnishing empty properties virtually using artificial intelligence.
Co-created by Mikhail Andreev and Michael Bonacina, both students at Harvard College and participants in the Harvard Innovation Lab, VirtualStagingAI claims to be the world’s first virtual staging platform powered by AI.
Mr Bonacina said the pair had used “several recently developed algorithms and machine learning models including DALL E, a version of GPT3 (which also powers ChatGPT) for images” to help develop the VirtualStagingAI.
While virtual staging applications have been in existence for some time, they typically require users to make decisions about where to place furniture and provide a limited selection of furniture to choose from.
Mr Bonacina said that VirtualStagingAI aims to save agents time by making these decisions for them, while removing the cost burden associated with having to physically furnish a property.
“Staging is proven to help real estate agents sell properties faster and for a higher price,” he said.
“Virtual staging is a lot more cost effective than physical staging and VirtualStagingAI is a lot cheaper than manual human virtual staging.”
Another benefit of using virtual staging is that agents aren’t limited to a certain catalogue of furniture.
“… in contrast to agencies the AI doesn’t work with a limited library of furniture pieces it can access which could lead to repetitive results but it instead dynamically generates furniture that suits the property and the desired style,” Mr Bonacina said.
The platform, currently offered for US$29 a month for unlimited use, simply requires an agent to upload a photo of an empty listing before using AI to offer a choice of different furnishing options.
It currently generates 10 different furnished rooms for the user to choose from.
“We output 10 different variations of mid-century modern furniture and as of right now usually two or three of them are good looking,” Mr Bonacina said.
“As the AI gets better and better with more data all 10 images should be viable options.”
The time it takes to produce the images (around 10 minutes when we tried it) will also reduce, he said, and different furnishing styles will be offered.
When Elite Agent agent trialed the technology it returned some impressive result, though there are a few pain points at this time, including the requirement that no doors be visible in the provided images to avoid a couch or table being awkwardly placed in front.
Beyond getting the program to output 10 perfect images each time, future product plans include expanding beyond living rooms and bedrooms to cover more areas of a listing, Mr Bonacina said.
He said that the platform is in its early stages and that agent feedback will be used to improve the product.
