SketchPro: AI Assistant for Architects and Designers

Given so much of the hype surrounding AI these days, it might be easy to dismiss the idea of an “AI assistant” for architectural design as one more example of a software tool that over-promises what it can actually deliver. However, it would be a mistake to do this for SketchPro, a brand-new AI application that lets you render and visualize a design in seconds starting from any sketch or 3D model. It goes far beyond what tools like Midjourney can do for architectural design (see the article, WA100 Live 2024 Webinar: Exploring the Impact of AI on Architecture, in which leading architectural firms described their use of Midjourney), and even AI add-ins to BIM applications like Archicad’s new AI Visualizer that allows high-quality visuals to be quickly created from conceptual massing models in Archicad (see the article, Graphisoft 2024 Product Launch).

I was recently introduced to SketchPro AI and had the opportunity to explore it in depth. I came away extremely impressed and amazed at what AI technology, as envisioned by the SketchPro team, has the power to do.

Note: The tool will be referred to as SketchPro in this review rather than SketchPro AI. It should not be confused with another tool called SketchPro, which is a generic sketching app rather than an AI tool for architectural design.

Overview

While SketchPro does include the familiar “text prompt” feature that is the mainstay of AI visualization applications like Midjourney, it also includes many additional AI-powered capabilities that allow it to be used as a design assistant to quickly flesh out design concepts and bring them to life. The starting point is an image, which can be a simple hand-drawn sketch, a CAD drawing, or a view of a 3D model. SketchPro works with this starting input to generate a large number of visuals in different styles, each of which can be further iterated upon until you hone in on one or more options that capture the design vision (Figure 1). The AI technology in SketchPro — which uses multiple machine-learning models, including both language models and vision models working together — enables this to happen almost instantaneously.

In addition to a text prompt to broadly describe the kind of rendering you want to apply to the input image, you can also select a “style reference” to guide the visual look and feel. This is one of the key strengths of the application, as it allows firms to compile style libraries of their previous renderings and use those to guide the creation of visuals for new projects in their “signature” styles. Style libraries can also be compiled of specific project types and/or in different regions, for example, office buildings in the US, university buildings in the UK, and so on. It is also possible for styles to be artistic — watercolor, charcoal, pen and ink sketches, etc. — to provide a more conceptual look and feel to the design (Figure 2). In fact, you can have entire libraries capturing, for example, different watercolor styles.

Other key capabilities of SketchPro include the ability to apply different materials to different parts of the image, insert assets like trees and people, delete objects in the scene, view the entire history of a rendering including all the iterations that were generated, see a side-by-side comparison of any iteration with the original input (Figure 3), and export a high-resolution file of the final rendering.

Let’s go ahead and dive into the details of the application.

How It Works

SketchPro is a cloud-based application, accessible through any browser, and does not require any local installation. Also, all the processing required to run the AI engine and generate the design options happens on an external server, so the computer being used to run the application does not require a dedicated graphics card or a high-end processor. Once you have an account, you can create multiple projects, and within each project, you can create multiple spaces, where each “space” is a design idea or concept that you are exploring. These spaces can be totally unrelated to each other, even if they are in the same project. Spaces can also be moved from one project to another, giving you the flexibility to organize the design ideas you are working on in any way you want. Figure 4 shows some examples of projects and spaces in two different accounts.

Importing and Rendering an Image

The starting point for a new space — essentially a new design idea you want to explore — is to import a base image in JPG or PNG format. This can be a hand-drawn sketch, a CAD drawing, or a view of a 3D model exported from any 3D or BIM application. The input image will typically have enough line work to provide SketchPro with an idea of what it represents. Once the image is imported, you can render the entire image using the different options in the Render tab (shown in Figure 5). These include a text prompt area where you provide a brief description of how the image should be rendered; for this, there is also an auto-describe option that can analyze the image and come up with a prompt that you can then modify as needed. You can set the level of creativity depending upon how much liberty you want SketchPro to take with the rendering; specify whether or not the original colors and textures in the input image should be retained in the output; and finally, select a reference image, either from the style library you have created earlier or by uploading a new reference image. As mentioned earlier, this gives SketchPro a visual example of what you are looking for.

Now that all the rendering options have been specified, you can click the Generate button to initiate the process. Based on the analysis of all the input parameters including the line work of the original image, the text description, the visual reference that was provided, and the creativity and color settings, SketchPro comes up with four to six different versions of the rendered output. The process only takes a few minutes, and the outputs that are generated are shown as thumbnails on the right, allowing you to browse through them (Figure 6).

At this point, if you are not satisfied with the results, you can choose to regenerate the rendering, making any changes to the input parameters if needed. Alternatively, you can select any of the outputs generated by SketchPro and use that as the next input to be rendered. In this manner, you can continue to iterate until you arrive at one or more design options that you are happy with (Figure 7). This ability to choose any output as the starting point for a new round of rendering — and keep this going until you arrive at what you are looking for — is where you can see the real power of AI. It would be impossible to do this using any other currently available technology.

Viewing the History

All of the different iterations created as you are working to hone in on the final output can be seen by going to the History tab. As shown in Figure 8, the entire history of the work on an image is captured by thumbnails in the left pane, showing the input image for each iteration and the outputs generated from it, with the most recent iteration listed at the top. Selecting any of the output images shows a preview of it, along with the settings that were specified for creating it including the text description and style reference.

For the output that is being previewed, a very useful feature is being able to compare it to the original image using a comparison slider. An example of this was shown earlier in Figure 3. Additional examples of this from different projects are shown in Figure 9. They illustrate how SketchPro retains the linework of the input image while applying the materiality and lighting of the reference image.

Refining the Output

Another important way in which SketchPro goes way beyond other AI visualizations tools is by being able to refine parts of the rendering output in different ways, all of which are specific to architecture and interior design. There is a Change Material tool, which allows you to apply or change materials only on parts of the image. The area to be changed can be specified by drawing it out or selecting it with a Photoshop-like Magic Wand tool. Once the area is selected, a palette with default materials opens up from which you can select the material that you want to apply to the area; you can also upload your own material or use a text prompt to describe a new material. The image will now be re-rendered with the new material applied to the selected area (Figure 10).

Additional capabilities for refining the output include being able to add assets such as people and trees to the image, populating it with entourage as is common in architectural illustrations. You can resize these assets and remove their backgrounds before placing them where required so they blend into the scene. Assets can be found by entering a word or phrase in the Search tab, and objects matching the description are pulled from a library of thousands of public assets across the web (Figure 11). You can add multiple assets to the rendering, controlling the order in which they should appear through a layering system. Once this is done, a Smart Enhance tool can be used to re-render the scene with all the assets you have added to it, adding, for example, shadows to the trees and people based on SketchPro’s analysis of the lighting in the scene.

Similar to how you can add assets to the rendering, you can also remove them from the image. SketchPro can remove objects leaving the rest of the scene untouched. This applies not only to objects you had placed but also other parts of the scene, such as any background objects that were generated by SketchPro based on the text prompt and the style reference for the render (Figure 12).

With all of these editing capabilities, you can continue to work on refining the output to your satisfaction. Once a rendering is finalized and you have used the Smart Enhance tool to render it in more detail, you can download it in high resolution to be able to use it in project documentation, client presentations, marketing material, and so on (Figure 13).

Conclusion

While I have an architectural background, I am not an artist or illustrator, and at least to my untrained eyes, the visualizations created by SketchPro seem amazing. And even though I have some understanding of how AI works with technologies like machine learning, neural networks, large language models, etc. (see the article, AI in AEC: An Introduction), I found SketchPro almost magical. I could have spent hours working with it — it was so much fun.

AI is fast becoming an industry of its own, with each day bringing news of more AI applications being developed for more use cases all over the world. For those working in the AEC industry, there is a lot of potential for AI to reduce inefficiencies, speed up workflows, reduce cost, and improve the quality of our built environment, ranging all the way from individual buildings and their interiors to roads, railways, bridges, and other infrastructure. I hope that tools like SketchPro are part of a new generation of AI tools for AEC that can turn the industry on its head and dramatically change how things are done.

About the Author

Lachmi Khemlani is founder and editor of AECbytes. She has a Ph.D. in Architecture from UC Berkeley, specializing in intelligent building modeling, and consults and writes on AEC technology. She can be reached at lachmi@aecbytes.com.

Blog

Have comments or feedback on this article? Visit its AECbytes blog posting to share them with other readers or see what others have to say.

AECbytes content should not be reproduced on any other website, blog, print publication, or newsletter without permission.