How did you make this?
Discussion
looks like automatic1111 (runs on stable diffusion), using img2img tab
- basic idea is you put the QR code into the img2img reference, then set the noise level to a degree that it will still pick up all of the data but give it a prompt for something completely different (like this might be "modern style, shleves, modern art, wall, " etc)
- i would set the noise level to try it out at like a 0.3 or 0.4. enough to change the image, but not enough to remove the data. You've have to tinker with it though.
Then just test it out and see if it scans. If it doesn't, increase the contrast or decrease the noise and keep experimenting. Remember a QR code actually has really good error correction built in, you can lose a good bit of the QR data while still properly reading it.
It’s also depends on the models! Gotta find out which one is a good fit.
Yeah I was playing around with it and getting pretty crappy results so far, lol. My initial guess on method isn’t very good 😆
Gonna do more exploring of models. I bet someone either has or may make a model or LoRA that works specifically for this task. Might try to find one that’s geared towards flat visuals or icon generation and see if it helps.