Turn any model view or image into an AI architectural render

A note from the creator

Fellow architects, interior designers, engineers, illustrators and more,

Having worked as a freelance architect for more than 2 decades, on both commercial projects and architectural competitions, I have been well aware of the effort it takes to produce photorealistic images to present your work.

Once upon a time, a single render would take hours, sometimes days to complete. Later on, rendering times vastly improved, but the price of professional software mostly remained well beyond what a young freelancer could comfortably afford. Recently, we have entered the brave new world of the ubiquitous (and honestly annoying) paid subscriptions, which was immediately followed by the rapid advance of the pervasive (and honestly exciting) AI.

While ray-tracing rendering is by far the most accurate and controllable way to produce high grade images, it requires a fair amount of workload: from setting up materials, lighting and cameras to multiple testing, adjustments and finally rendering. AI, on the other hand, though a bit unpredictable, is effortless, fast and produces easily impressive illustrations. However, if you do not have the time and determination to learn a whole new set of yet complicated tools, your alternative is to surrender your work to one of the establishments (:-) that provide the service online, while also having to keep that little subscription anxiety in the back of your mind.

That's why I decided to build THETA Render.

THETA Render is an indie software project that packs open-source diffusion models and a set of pre-compiled prompts into a minimal app that will turn your images into photorealistic illustrations. As a sideline it can enlarge the generated or your traditionally rendered images (or any image for that matter) to resolutions up to 40 or 50 megapixels. THETA runs locally on your own NVIDIA RTX GPU, so it involves no daily limits, no cloud rendering on an online AI service, and no sending your images or projects to me or some corporate entity. Your work never leaves your PC.

It's fast, straightforward to use, and does not expect you to become an AI expert just to get an image out of it. Mind you though, as diffusion models can occasionally invent details, alter elements or misunderstand parts of your design, you may need to hit the retry button a couple of times. Still, when you want a strong and convincing image quickly, THETA Render can be a very useful addition to your workflow.

To make things effortless, I have included SketchUp and Rhino 3D plug-ins (installable directly through the app), which send the active viewport to get rendered with the click of a button. You're not just limited to those programs however, since you can also load, paste or simply drag and drop image files right into THETA Render.

I've tested it on a small range of hardware. From a laptop RTX 2060 with 6 GB of VRAM to a desktop RTX 4070 with 12 GB and a 4080 with 16 GB, the results vary from acceptable to highly impressive. However, as I don’t have access to other advanced GPUs, you can test the trial version on your own rig before deciding to buy.

THETA is a paid app as I’d honestly like to make some additional income from this :-). But instead of yet another subscription, the deal is simple: you pay once and keep it. No monthly fee, no expiry, no account. The lump-sum cost is intended to amount to about what a month or two of an AI subscription does.

THETA Render is an independent project, made by an architect who wanted a practical, affordable and private way to create better images. It is not perfect but is fast, attractive and effective. Give it a try and I hope it will prove useful to you, too.