Free AI Architecture: From Sketch to Stable Render
Are you on the hunt for an easy yet effective way to bring your architectural sketches to life? Imagine transforming those sketches into stunning, realistic renders almost as if you're seeing your designs come to life before your eyes! Whether you're a seasoned architect or a student mastering the art of architecture, finding the right AI tool to make this magic happen is crucial. And guess what? You've landed in the perfect spot! In today's article, we're diving into the world of Stable Diffusion Automatic 1111, a top-notch image generative AI tool. I'll guide you step by step on how to install this program, load the essential checkpoint models, and utilize the incredible Control Net extension. This powerful feature lets you steer the output to match exactly what you've envisioned.
Setting up Stable Diffusion Automatic 1111
Now, let's kick things off by setting up Stable Diffusion Automatic 1111. First things first, to run this AI program locally on your computer, there are a few requirements you need to meet. An Nvidia graphics card is highly recommended for almost all generative AI programs. To install Automatic 1111, you can head over to its GitHub page (link in description) and scroll down to find detailed installation steps. If you find this process a bit complicated, I've previously made a video that walks you through the installation, which you can find linked right here. For a more user-friendly installation option, you might want to check out a software called Pinocchio. It simplifies installing Automatic 1111 along with other top AI tools without the need for any coding or hassle. I have covered how to use Pinocchio in another video, which you can also find linked here.
Getting Started with Your Project
Now, let's move on to the exciting part - getting started with your project! Begin by visiting the Civit AI website to download a checkpoint model that's specially trained for architecture. This model is fantastic for creating amazing exterior images. Before we dive deeper, make sure to download the SD 1.5 base models, as today's tutorial is specifically focused on working with this type of checkpoint model.
Once you're on the Civit AI website, take a moment to scroll down. There, you'll find all the essential information about the model, including detailed instructions on how to use it. An exciting feature you'll discover is a gallery of images created with this base model by other Civit AI users. These images not only showcase the model's capabilities but also provide insights into how each image was crafted. We'll delve more into this a bit later, but for now, let's focus on downloading the checkpoint model. To download it, simply click on the download button. When saving the file, navigate to the Stable Diffusion web UI folder on your computer. Within this folder, go to the models directory, then find the Stable Diffusion folder and save your file right there.
Now, it's time to launch Stable Diffusion Automatic 1111. To do this, simply click on the web_ui.bat file and wait for the web user interface to open in your browser. Once it's up and running, you're all set to start transforming sketches into stunning realities!
Generating Realistic Renders
First up, click on the checkpoint model refresh icon button. In the drop-down menu that appears, select the model we downloaded, which is Architecture Exterior v90. Now, we're going to use the Control Net tab. Here, drop your sketch image into the designated area. Keep in mind that your sketch should ideally be 768 x 512 pixels. We're using this resolution because most SD 1.5 models are trained on images of this size, making it optimal for our purposes.
After you've uploaded your sketch, choose Line Art as your control type model. Remember to tick the Enable and Pixel Perfect checkboxes. Now, adjust the width to 768 and change the sampling method. Increase the sampling steps to 30. This will enhance the detail in your generated images. Set the batch size to two, allowing us to generate a pair of images to choose from.
For the positive prompt, take inspiration from the images on Civit AI that were created using this model. Describe the desired appearance of the house, including materials, environment, lighting, and other aspects. Conversely, use the negative prompt to specify what you don't want in the image. As always, all the essential links and prompts can be found in the description below. When you're ready, click on Generate to watch as your sketches come to life in two unique renditions!
Take a look at the two images we've just generated. They're absolutely stunning! While there may be slight differences between them, like variations in colors and lighting, the key takeaway is how well Stable Diffusion followed our sketch input in Control Net and adhered to our prompt instructions to deliver these results. Remember, you can always hit Generate again to explore more options or tweak your positive and negative prompts for even finer results.
Refining Your Images
In this example, I'm really impressed with the lighting and overall feel of one image, but there's a small issue - the AI added birds in the sky, which we don't want. Let's fix that by using the inpainting feature. Just click on the color palette icon and you'll be directed to the Imageo Image tab. Here, you can mask the parts you want to change. Make sure to clear out our previous prompts and replace them with specific instructions for the adjustments.
In our case, to remove the birds, we'll type "Clear Blue Sky." Adjust the only MK padding pixels to around 60 and play around with the denoising strength, starting with a higher value and adjusting until you get the desired result. And there you go - the birds are now gone! To further enhance our image, let's upscale it using the Ultimate SD Upscale extension. You can install it from the Extensions tab. Click Install and wait until a message confirms that the extension is installed.
You'll need to restart Automatic 1111 by closing the CMD window and starting the web UI again. Then, you're ready to use it in the Image to Image tab. Set the DOOA strength to 0.2, then select Ultimate SD Upscale and opt for the 4 Ultra Sharp Upscaler. Make sure to adjust the scale according to your needs. In my case, I want to upscale the image twice its size. Click Generate to upscale your image, and notice how it becomes more detailed and higher in quality, closely matching our original sketch's real-life appearance.
This workflow is applicable to any sketch. Just replace the sketch on Control Net and adjust the prompts. Remember, you can emphasize certain details by enclosing them in brackets. For example, if you want a red-tiled roof, just specify it in your prompt. The result? A beautiful house with the exact roof you envisioned!
You can then upscale the image for an even more impressive outcome. As you become more experienced, feel free to experiment with different checkpoint models and Control Net types. The beauty of Automatic 1111 is its versatility. And best of all, it's completely free! No subscriptions or fees.
Conclusion
That's all for today's tutorial! If you enjoyed this article, please give it a like, share it, and don't forget to subscribe if you haven't yet. For those of you who are AI enthusiasts or interested in learning how to monetize AI tools, check out my other videos. I cover a wide range of AI-related topics on this channel. Thank you for reading and happy sketch-to-render transformation!
FAQs
Q1: Are there any specific requirements to run Stable Diffusion Automatic 1111?
A1: Yes, it is highly recommended to have an Nvidia graphics card for optimal performance with most generative AI programs, including Stable Diffusion Automatic 1111.
Q2: Can I install Stable Diffusion Automatic 1111 without any coding?
A2: Absolutely! There is a user-friendly software called Pinocchio that simplifies the installation process for Stable Diffusion Automatic 1111 and other top AI tools. You can find a tutorial on how to use Pinocchio in one of my videos.
Q3: How can I enhance the details in my generated images?
A3: To enhance the details in your generated images, you can adjust the sampling steps to a higher value, such as 30. This will result in more detailed and refined renders.
Q4: Can I remove unwanted elements from my generated images?
A4: Yes! In Stable Diffusion Automatic 1111, you can use the inpainting feature to remove unwanted elements from your generated images. Simply mask the parts you want to change and provide specific instructions for the adjustments.
Q5: Is there a way to upscale my generated images for a higher quality outcome?
A5: Absolutely! With the Ultimate SD Upscale extension, you can upscale your generated images for a more impressive outcome. Just make sure to adjust the scale according to your needs and enjoy the higher quality renders!
