Imagining the Home of the Future with Generative AI
Introduction
It’s midway through 2023 and you’re reading yet another blog post about Generative AI. By now you’ve probably already used, seen, or heard about Generative AI through OpenAI’s ChatGPT or viral images like Pope Francis wearing an oversized puffer jacket. At Homebound, one of our core values is “To Build Better” by rethinking every step of the home construction process and defining the future of homebuilding. That’s why we began exploring how AI could improve our existing processes and be used to better communicate with our customers in the future. From designing homes to generating personalized content, we believe there are many ways Generative AI will disrupt the real-estate industry.
End-to-End Home Design
The first, and probably most obvious way we see image generation changing the industry is through design. Using designer input and domain knowledge, homes could be designed end-to-end from creating a floor plan, to manufacturing an exterior, and ultimately an interior.
Below, we prompted (the process of using text to guide AI to output what you want) Midjourney to design an entire home inspired by Homebound’s product line. The relevant prompts used to generate each image are located in their respective captions.
Floor plan
Starting with the floor plan, we wanted to create a single story home containing a few bedrooms, a living room, a kitchen, and a garage. While the floor plans generated didn’t deliver on everything we wanted, we loosely followed the generated floor plan below to design our interior and exterior.
We then sketched this floor plan out to have a clearer picture of what we wanted our house to look like when designing it.
Exterior
Once we were happy with our floor plan, we created an exterior modeled after Homebound’s Modern Farmhouse product line.
Interior
Lastly, we generated the interior of our imaginary home. Beginning with the kitchen, we created communal spaces with tons of natural light and modern finishings that bled into the living room.
For the primary suite, we elected to go with earth tones and green accents for both the bedroom and bath.
The other two rooms were filled by adding a secondary bedroom and office. Each room got its own unique twist following the core design of the rest of the home.
Mood Board & Color Palettes
While we used Midjourney above to design entire rooms, it can also be leveraged for smaller pieces within the process. Even when explicitly prompting for what you want, generative models will ultimately take creative liberties on certain aspects of an image. Below are two examples of a mood board and color palette generated for kitchen (left) and bathroom (right) fixtures.
Personalization
One way Homebound is already improving the home buying process is through our personalized property checkout workflow. This allows customers to view Homebound properties under construction and personalize them to their liking. Exteriors, materials, layouts, and more are all configurable through our checkout process.
Using Generative AI, we can imagine an infinite number of variations of the same home, design, etc. that can inspire both customers and designers in the build process. Taking our generated home exterior as an example, we’ve created numerous looks as similar or dissimilar to the original design as we wanted.
These models can be also be used to modify existing renderings and images. Using an open-sourced model hosted on Hugging Face, InstructPix2Pix, we modified a Homebound home’s kitchen rendering, turning the original navy blue kitchen cabinets (left) to dark brown (right). Using prompts similar to the one in the caption below, customers could quickly visualize homes in any color, material, or style they dream up. Additionally, designers could use these models to rapidly test looks they’ve imagined. Swapping out furniture, virtually staging rooms, etc. are all possible with these types of models.
Looking Forward
As we’ve demonstrated above, generative AI can be a powerful design tool for the home building process and better bridge the gap between customer and builder interaction. While we’ve highlighted a few use cases through Midjourney and other open-source models, we believe these models will begin to become integrated in existing tools such as with Adobe Photoshop’s generative fill and also create new ones.
Companies like Oda Studio are using AI to virtually stage rooms to allow potential home buyers to see what a space would look like with furniture. Engineering start-ups such as Hypar are creating generative Text-to-BIM (Building Information Modeling) platforms to allow engineers to quickly design and develop projects. Even real-estate companies like Redfin are integrating ChatGPT plugins into their platforms to help people find homes quicker. While generative AI is still evolving at a rapid pace, it’ll be interesting to watch how this technology will improve the construction and real-estate industry in the years to come!
Homebound is hiring! Interested in working at Homebound? Check out our careers page!
To learn more about Homebound, visit the Homebound website or go over to the Homebound Technology Blog to view more stories from our team.
References
Here are some of our favorite AI content creators and resources that inspired content for this blog:
[1] Most prompts above follow the additive prompting pattern outlined by Nick St. Pierre on Twitter.
[2] Stephen Coorlas on Youtube for generative AI use cases in architecture.
[3] Midjourney Command List for learning more about prompting commands.