(Work in Progress)
Expert Guide for QR Code AI Art with Stable Diffusion
In this document we will discuss the "Expert" options in the QR Diffusion Generator. Most of them should be used with care, as the default options should be sufficient for the average user.
QR Options
There is just one new setting in Expert mode.
Error Correction
The QR code error correction is a percentage that allows QR codes to be some parts of it are covered or smudged and yet still work just fine.
Normally we recommend "H - High (30%)" level of error correction, but in some cases.
If you have long QR code data and QR code is complex, lowering the error correction makes the QR code simpler in shape. The AI QR codes based on the simpler input QR code will be more creative and easier to read.
ControlNet Options
Model
In the expert mode you can select the ControlNet model which is responsible for baking the QR code into the image while generating. There are number of models available, while the default yields the most stable results.
- QR Code by Dion Timmer (default)
- QR Pattern V2 by Nacholmo
- QR Code by Monster Labs
QR Guidance Range
Select a start and end point of when the ControlNet is active in the process of generating the image. ControlNet is baking the QR code in the image being generated. Normally you want it to be active during most of the image-generation process, but sometimes you can experiement and have the generator ignore the ControlNet for either the start or end of the generation process.
Stable Diffusion Options
Sampler Method
[TODO]
CFG Scale
CFG scale (classifier-free guidance scale) or guidance scale is a parameter that controls how much the image generation process follows the text prompt. The higher the value, the more the image looks like the given text input.
Denoising Strength
Lower Denoising Strength will make your generated images look similar to your input image.
Higher Denoising Strength increases variation and reduces the influence of your input image on your output image.
Seed
Seed is an integer number that when entered, it produces similar images for the same seed. When no seed is entered, the generator chooses a random seed.
It does not work the same way when ControlNet is involved, so this setting does not have a real effect in our case and it is not really relevant for QR code generation with ControlNet.