As we are all aware, AI has become a constant news headline. In fact, as I’m writing this newsletter, Amazon announced Rufus, a new generative AI-powered conversation shopping experience. What will the headline be tomorrow?
This week, we’re taking a look at how we can begin to unlock AI’s potential and start integrating it into our presentation palette. The images and descriptions below provide a snapshot of our AI introduction.
Image #1 - The process begins with a volumetric BIM model with a moderate amount of detail. Materials are graphically represented by solid colors allowing easier modification during the next step.
Image #2 - Once the basic model contains the necessary information we export it to a real-time visualization tool that links directly to our BIM software. This visualization tool allows us to quickly add elements like textures or materials, lighting, location, and vegetation.
Image #3 - Images generated by the real-time visualization tool are imported into the AI software. This software allows us to input positive and negative prompts that describe our parameters like style of architecture, location, time of day, materials, weather, etc. After we’ve completed inputting the prompts, the software allows us to control other variables such as geometric freedom (the amount of deviation from the original image), the type of image we desire (watercolor, sketch, photorealistic), and output image size. This software also allows us to highlight areas of the rendering for further iterations which is extremely useful.
So what are our thoughts about implementing AI into our tools? It may be beneficial for us to incorporate AI technology into our tools, as it has become increasingly relevant in our field and could have a positive impact. Proper utilization of the software is essential, particularly for those who are new to it. During our research phase, we discovered that images with tighter vignettes produced better rendering results than those with overall compositions. It is also worth noting that some post-production Photoshop work may be required. While AI may not always be perfect, we are pleased with the overall outcomes and will continue to explore its potential.