ControlNet Breakdown (Diffusion Models For Architectural Renderings)

Welcome to this week’s edition of Architecture Insights.

AI image generators have several advantages such as being user-friendly, generating images quickly, and aiding conceptual ideas. Conversely, manual rendering tools such as Photoshop offer full control over the image creation process. However, both methods have their respective restrictions.

ControlNet is the closest tool to provide the benefits of both with minimal drawbacks.

AI News For Architects & Designers

  • Midjourney brings the Zoom & Pan & Vary feature to V6.

Summary: Zoom, pan, and vary are common prompts in Midjourney and if you’re a frequent user of version 6 then these additions are useful. Also, type /info into Midjourney to see if you have over 5000 generated images, if so you are now able to access Midjourneys alpha website.


What is ControlNet?

ControlNet is a neural network model that introduces an additional layer of conditioning, In simple terms, it is an addition to stable diffusion models that offers more parameters to images.

It is designed to give users more control over the output of Stable Diffusion (a common type of AI image generating software) models. This is useful for architectural renderings because we have more control over what our images look like in the end.

Consider this comparison: Midjourney provides excellent AI image quality and concepts but lacks customization features. Photoshop, on the other hand, requires a lot of manual user controls to create high-quality renderings (until the recent introduction of AI features such as generative fill). ControlNet aims to combine the best of both.

ControlNet Advantages Vs Other AI Image Generators

A few standout features:

Depth Maps for Enhanced Spatial Awareness

ControlNet utilizes depth maps to define the spatial features of images. You can adjust the depth field and perspective in AI-generated visuals, thereby improving the realism and spatial accuracy of the output.

Image Revitalization

Enhance and repair historical photos or damaged visuals with some of the highest levels of accuracy in comparison to other image generators.

Edge Detection

Used to locate and highlight significant edges in images, letting you make your renderings accurately target and modify individual components with high precision.

Stable Diffusion & Automatic1111

A common way to get started with ControlNet is through Automatic1111, a tool that gives you access to stable diffusion through the web. Stable Diffusion’s true potential in architectural design is unlocked through web interfaces like Automatic1111 which simplify the interaction with the model.

You can also get access to a preview of the full version.


It can be challenging at the beginning to set up and run the tools mentioned above, from technical aspects of stable diffusion models to the ways of using ControlNets features. To assist you, we have a list of valuable resources.

Platforms for exploring open-source stable diffusion models: Hugging Face & Civitai

How to Install Automatic1111 on Mac or Windows: Windows & Mac

How to Install ControlNet on Automatic1111: Install ControlNet

2-minute workflow turning sketch into an image on ControlNet: Video

An extensive guide to using Automatic1111, the stable diffusion WebUI: Guide

In the upcoming weeks, we will be sharing short and informative tutorials on how to use the various features available in ControlNet models.

Some features to go over:

  • How to use ControlNet to turn rough sketches and line art into high-quality renderings.

  • How to use positive and negative prompts for your images.

  • The different types of parameters in ControlNet.

  • Comparisons to other image generators.

AI Image of the week

Thank you for reading this week’s issue, check past issues here. Share this newsletter with colleagues, friends, or anyone interested in the combined world of architecture and artificial intelligence.

Feedback is always appreciated, let us know how we did by responding to this email.

Until next Friday,


Did you enjoy this week's post?

Login or Subscribe to participate in polls.