As various industries increasingly adopt Artificial Intelligence, have you considered its role in interior design? While you may be familiar with Midjourney's AI-generated interior design, there are other AI tools available for this purpose that you may not have heard of yet. Today, you'll discover something new.
Starting from Archicad 27 an experimental AI Visualizer add-on, powered by Stable Diffusion, available for both Mac and PC. Find more details 👉here.
This article showcases two tools that can be utilized to generate various representations of your BIM model through renderings. And even if you think that all the renderers will be out of a job, it is too early to say that AI will replace them.
Let's begin by introducing the tools we use: Collov GPT and Room GPT. Here's a brief 15-second overview of the basic workflow for both:
👉 Interior Design AI Generator Collov GPT: gpt.collov.com
👉 Dream Room Generator RoomGPT: roomgpt.io
Here are some our tips
After experimenting with both RoomGPT and Collov.GPT, we found that RoomGPT gave us better results. The final picture in Collov.GPT tends to be blurry and less predictable. Therefore, we recommend using RoomGPT.
What inputs can you use?
Designing your project in BIM, such as Archicad, allows for exporting multiple model views with ease. Consider using the following model representations as input:
3D documents
Elevation views (yes, it works from time to time!)
Renders with textures or white surfaces
3D window screenshots with 3D Styles as 3D window Style or Simple Shading with contours
Inputs insights
While AI engines are powerful tools, the input picture can also play a significant role in the quality of the output. Here are some Do's and Don'ts based on our experience:
Don'ts
Don't use renders that are too dark or overexposed, as this can negatively impact the AI's ability to describe the object or scene accurately;
Avoid using images that are too cluttered or busy, as this can make it difficult for the AI to focus on the object or scene you want to describe;
Avoid using pictures that only show a partial view of the furniture, as this may lead to unexpected interpretations by the AI;
Avoid using model cuts, isometric views, and top views, as these services are only capable of creating standard perspective views at the moment.
Currently, the AI doesn't seem to differentiate much between textures and surface fills, so don't put too much emphasis on them for now.
AI is still in its early stages, so don't expect incredible results just yet. It's best to view it as an additional tool, rather than a complete replacement for traditional renderings.
Do's
Select the appropriate category for the room
Use perspective views for the input picture
Ensure that the object or scene is well-defined and clearly visible in the picture
Use high-quality images with good lighting to assist the AI in accurately interpreting the object or scene
Consider using contours in the 3D model to aid in interpreting the object
Avoid cluttering the scene with too much furniture, as this can distract the AI and lead to less predictable results
Experiment with multiple tries to achieve the desired output.
Bonus!
We stumbled upon a new tool for interior design floor plans that generates furniture PNGs with different fills. While the results may be unpredictable, it's still worth experimenting with for some fun and creative design options.