Skip to content

Beyond Ugliness: Reimagining QR Codes with AI

QR codes are everywhere these days. For the longest time, people resisted using them because they looked ugly, you needed to download a separate app to scan them, and frankly, they were a hassle. But since the pandemic, QR codes have seen a major revival. Plus, we’re starting to see a lot of creative hacks to make QR codes look good and blend well with solid design and illustrations.

00467 3396652256 oil painting of a scottish village white background


Creating good-looking custom QR codes manually can take a lot of trial and error, and a lot of time. You also need to know the nuts and bolts of how a QR code actually works. But thanks to AI, all of this can be simplified. You can take an existing image, or just prompt anything you can dream up, and merge it with your QR code.

Here are a few more QR codes I’ve made with AI:

00464 928286874 painting of a sri lankan village white background

00458 1199888871 vector logo of a lion illustration clean lines black and white white background

00470 3746435178 oil painting of a European town

00477 3829944748 watercolour painting beach bahamas ocean palm trees pastel colours

Here are the details and setting I used to make my AI-generated QR codes:

The Settings I used within Stable Diffusion and ControlNet are below

I personally use Vlad Diffusion (SD.Next), but it will work with Automatic 1111 too.

  • ‣ Steps: 22
  • ‣ Sampler: DPM++ 2M Karras
  • ‣ CFG scale: 6, Seed: 1597248777
  • ‣ Size: 512×512 or 768×768 (this can be upscaled later)
  • ‣ Model: dreamshaper_631BakedVae
  • ‣ Clip skip: 1,
  • ‣ ControlNet 0:
    • ‣ ‣ preprocessor: inpaint_global_harmonious
    • ‣ ‣ model: control_v1p_sd15_brightness
    • ‣ ‣ weight: 0.35
    • ‣ ‣ starting/ending: (0, 1)
  • ‣ ControlNet 1:
    • ‣ ‣ preprocessor: inpaint_global_harmonious
    • ‣ ‣ model: control_v11f1e_sd15_tile
    • ‣ ‣ weight: 0.65
    • ‣ ‣ starting/ending: (0.21, 0.71)