TheFrontEnd🔥

#personal #uiux #learn

All of the above are also valid QR codes. Some people asked me to share more on that - there you go then 😂

First of all this images are from qrbtf - a service that attempts to automate this whole process. You can jump in and try to generate yours quickly with some prompts. The only thing it won't let you change url the QR points to as the whole system is still WIP.

But all they do under the hood is use stable diffusion with specific controlNet that integrates your QR into generated image. I really recommend trying it yourself though as the process is quite fun and you learn along the way. And I'll share all the links you need to get started.

There is one specific guy that deserves all kudos in the world that wrote the whole 101 on it. Not gonna reinvent the wheel here - so https://antfu.me/posts/ai-qrcode-101

I managed to even get it running on my lower M1 mac, so you should be able too.

The quickest way to get started

- install stable diffusion web ui
- get the controlnet extension
- now get controlnet qr pattern

And you are pretty much set. Then just follow tutorial from antfu.me and get your qr codes flowing in.

I personally also found useful to get revanimated - the quality it can generate even with simplest prompts is phenomenal.

The best sampling methods that I liked were DPM++ 2M Karras and DDIM.

Now to the most interesting part 🔥

It's all gonna come down to "oh, I cant scan the code 😭" vs "well fuck, it's obvious its a QR code 🫣". And you ll have to play with the control net settings.

Also one thing I didn't know is that not all QR scanners are the same, with this QR codes some might not be good enough. As a rule of thumb, try different ones - eg wechat scanner or iphone (there are 2, in camera and dedicated one).

And because you read this channel you'd already know that the most important things for any scanner is the finder patterns (those 3 large squares) - they tell the scanner it's actually a QR code so scan it pls.

Then smaller square is alignment pattern - becomes super important if you rotate you embedded QR code. Get this 2 points right and you ll increase the scannability.

Finally error correction is important, make sure it's not the lowest (aka you need to have more elements in the QR itself) cuz some of those can get missed when generating the image.

You can also completely change the way you input QR code looks, circles vs squares, dots vs lines etc. By playing with those you get more natural integration into objects in the images.

Like the one I had for this channel doesnt have obvious square finder patterns but you can still spot the elements in the table drawers that are pretty much those patterns. It reduced scannability for some scanners as a side effect but still made me happy and was good for POC.

That should be just about the largest post this year ever 😀 and just about enough info for you to try it yourself if you also find it amazing like I do.

Qrbtf

Gallery - QRBTF

QRBTF AI QR Code Generator

www.tgoop.com/thefrontend/1052

9.1K viewsAug 4, 2023 at 07:30

#personal #uiux #learn

All of the above are also valid QR codes. Some people asked me to share more on that - there you go then 😂

First of all this images are from qrbtf - a service that attempts to automate this whole process. You can jump in and try to generate yours quickly with some prompts. The only thing it won't let you change url the QR points to as the whole system is still WIP.

But all they do under the hood is use stable diffusion with specific controlNet that integrates your QR into generated image. I really recommend trying it yourself though as the process is quite fun and you learn along the way. And I'll share all the links you need to get started.

There is one specific guy that deserves all kudos in the world that wrote the whole 101 on it. Not gonna reinvent the wheel here - so https://antfu.me/posts/ai-qrcode-101

I managed to even get it running on my lower M1 mac, so you should be able too.

The quickest way to get started

- install stable diffusion web ui
- get the controlnet extension
- now get controlnet qr pattern

And you are pretty much set. Then just follow tutorial from antfu.me and get your qr codes flowing in.

I personally also found useful to get revanimated - the quality it can generate even with simplest prompts is phenomenal.

The best sampling methods that I liked were DPM++ 2M Karras and DDIM.

Now to the most interesting part 🔥

It's all gonna come down to "oh, I cant scan the code 😭" vs "well fuck, it's obvious its a QR code 🫣". And you ll have to play with the control net settings.

Also one thing I didn't know is that not all QR scanners are the same, with this QR codes some might not be good enough. As a rule of thumb, try different ones - eg wechat scanner or iphone (there are 2, in camera and dedicated one).

And because you read this channel you'd already know that the most important things for any scanner is the finder patterns (those 3 large squares) - they tell the scanner it's actually a QR code so scan it pls.

Then smaller square is alignment pattern - becomes super important if you rotate you embedded QR code. Get this 2 points right and you ll increase the scannability.

Finally error correction is important, make sure it's not the lowest (aka you need to have more elements in the QR itself) cuz some of those can get missed when generating the image.

You can also completely change the way you input QR code looks, circles vs squares, dots vs lines etc. By playing with those you get more natural integration into objects in the images.

Like the one I had for this channel doesnt have obvious square finder patterns but you can still spot the elements in the table drawers that are pretty much those patterns. It reduced scannability for some scanners as a side effect but still made me happy and was good for POC.

That should be just about the largest post this year ever 😀 and just about enough info for you to try it yourself if you also find it amazing like I do.

BY TheFrontEnd🔥