Everyone who is an active user of any AI art generator knows the main issue.
The generated images are too small. I’m talking about 1024 x 1024 pixels, which is the maximal output these days.
And if you are willing to sell AI art on any print-on-demand marketplace, you need to upscale these images.
For example, the recommended dimension on Redbubble is 6000 x 8000 pixels or 6000 x 6000 pixels.
I want to share with you one credible tool that can upscale your Stable Diffusion artwork and even fix faces. And this AI image upscaler is free and open source.
Free AI Upscaler That Fixes Faces
This is a project on a Google Colab, and it can do magic for your designs. It is called Real-ESRGAN and it can upscale any image, and enhance it to insane results.
And I have tested it on artworks with faces that have been generated with AI, and you know that AI has still issues with faces.
And I have been surprised by how the Real-ESRGAN can fix or improve faces on most images. In fact, this ai tool helps me upscale images for free and open source.
For better visualization, I have recorded a video tutorial on how to use the Real-ESRGAN model. You can watch it below.
To run this Colab, you need to be logged in to your Google account. Once you’re logged in, you need to run all the operations needed.
Just click the play icon in the Preparation section, after it is done, you go click the play icon for the Upload Images and upload the image you want to upscale and improve by clicking the choose file button.
In the next step, you should run the Inference code block, but before you do that, if you want, you should change how much to upscale the image. Just change the value next to outscale command to a value you want.
You can also disable the face enhance feature, by deleting this part of the code.
This is optional, you can use the Real-ESRGAN model with the default settings.
Now, it is time to play the Inference block code and then run the Visualization block.
Sometimes you need to wait a bit longer to get the result, but in the end, your image should be hugely improved.
As you can see, there is a huge difference in faces in these images.
Finally, you should download the results by running the Download block, but if this doesn’t work, just click on the File icon on the left side, open the Real-ESRGAN folder, and then open the Results folder.
This is a folder, where your enhanced image is stored.
As I already mentioned, using Real-ESRGAN is easy, straightforward, and can be used by anyone who needs to upscale and improve artwork made by AI and sell it on any print-on-demand marketplaces.
I hope this article was helpful, and if you have any questions about AI art, Stable Diffusion, or print-on-demand, you can use the comment section, or read more about all of these topics on this website.
BONUS FOR YOU: Get the Digital Product Starter Kit and start building your own online business today.
Here Are Some of My Favorite Tools For Print-on-demand
As a graphic designer and POD seller, I’ve had the opportunity to discover different helpful products and tools that can be time-saving and make the process of creating your designs a bit smoother.
DESIGNS: Kittl (best for t-shirt designs), Vexels (for professional designs), Placeit (for unique product mockups)
AI IMAGE GENERATOR: Leonardo (best free account)
GRAPHICS: Creative Fabrica (cheapest marketplace), Envato Elements (more variety)
SELLING DESIGNS: Freepik (for beginners), Creative Fabrica (for advanced graphic designers)
SELLING LOW-CONTENT BOOKS – Book Bolt (budget option)
ETSY:
- Research – Alura (best), EverBee ( beginners), Sale Samurai (cheap)
- Fulfillment – Printul (beginners), or Printify (advanced).
Disclosure: Some of the links above may contain affiliate partnerships, meaning, at no additional cost to you, NechEmpire may earn a commission if you click through to make a purchase.
- Simplify Your Etsy Shop: Fast Listing Tips for Everyone - November 5, 2024
- The 4 Etsy Essentials: Tips for a Strong Start in 2025 - October 13, 2024
- RGB vs. CMYK: One Reason Why Your Designs Look Different on Screen and in Print - October 9, 2024