So in short, the past six months have been quite an adventure, left my day job, started a business, and moving to AR to start my own path as an architect-developer. In the interim, I have been exploring the use of AI for my day-to-day practices and thought would be nice to share with the community. My previous post was about GPTs, now would love to chat about Stable Diffusion Models and Image generation for interior/exterior inspiration.

I use mainly three different AI vendors, which are Midjourney, Dall-e & Stable Diffusion. While the first two are great for inspiration, the last one offers much more image control based on what is called ControlNets, which allows control of the composition and weights of an image. It can read an image and apply a new image within context. For example, let's stage an empty room, remodel a facade, do a new bathroom idea etc.

There are three main workflows I find promising so far, sketch to image, model to image and image to image which I will describe below:

Sketch to Image - By using a local installation of Stable Diffusion, you can create image generation workflows and adapt them for your uses. For example an architecture style, graphic style, exterior vs interior and so forth. It requires technical knowledge yet is super adaptive and open source and once controlled, you can easily scan sketches of ideas and convert them to realistic renderings.

Model to Image - Some apps such as Veras connect to your Sketchup/Rhino/Revit software and can directly render simple masses into final results. Easy to install but requires you to have prior design software knowledge.

Image to Image - There are many AI apps out there online that can modify images. Won't market a specific one (Feel free to reach out privately), but those are easy to use, nontechnical, and provide okay + results. I find it quite useful with clients, or just trying to envision ideas on potential properties.

All of them are based on ControlNets and StableDiffusion, so if you have a bit of tech savviness I would recommend testing it out locally and start exploring it for your uses. For direct SD installation, you can use Automatic1111 here:https://github.com/AUTOMATIC1111/stable-diffusion-webui or use web native apps like Rendair to test it out on the web without much of a commitment.

I've attached a sample of work from an existing image from Zillow, and an AI rendering as an example. Feel free to connect for questions and would love to learn how others are using AI in their day-to-day business operations.

Have a great day!