[Stable Diffusion] improves illustrations quality with LoRA: "flat2"

  • What kind of effect does flat/flat2 have?
  • If you can easily improve the quality of the illustration, maybe I should use it...
We will answer these questions for you.

Stable Diffusion LoRA flat2

Recently, there has been a lot of talk about Stable Diffusion LoRA, which is called flat, and that it "improves the quality of illustrations." If the effect is low, it's just a waste of time to introduce it.

Stable Diffusion LoRA flat2
Stable Diffusion LoRA flat2
In this article, I will explain about :
  • [Stable Diffusion] Introducing LoRA: "flat2" that easily improves the quality of illustrations
  • How to download and install flat2
  • If you want to generate illustrations more smoothly...
  • How to easily manage a large number of generated illustrations
so if you read to the end, you will understand how to use flat to improve the quality of your illustrations.

What is LoRA in the first place? If you don't know how to use it, please read the following article first :

>> What is LoRA that can be used with Stable Diffusion? Explain how to install and what to watch out for!

Introducing LoRA: "flat2" easily improves the quality of illustrations

flat2 is LoRA that makes a "flat picture" as the name suggests :

Introducing LoRA: "flat2" easily improves the quality of illustrations
Introducing LoRA: "flat2" easily improves the quality of illustrations

Originally for this illustration :


If we apply flat2 with weight 1 :

LoRA: "flat2"
LoRA: "flat2" 

It will change to a flat illustration with a little drawing. It's LoRA in the direction of "reducing quality".

turn this around :
  • Prompt field: Apply flat2 with weight -1.
  • Negative prompt field: apply flat2 with weight 1.
By inputting in this way, the idea is to increase the amount of drawing and improve the quality. Be aware of the negative weights in the prompt field.


When I actually generate an illustration under this condition.

flat2 LORA
flat2 LORA

It's obvious that the amount of drawing is increasing. it's amazing…. If you pay attention to the hair in particular, the difference will be easy to understand.

I tried the same with "flat", the predecessor of flat2.

Stable Diffusion LoRA
Stable Diffusion LoRA

I feel that the amount of drawing is not much different from flat2, but the coloring is a little different. Personally, I prefer flat 2, so I'm thinking of using it mainly.

This time I tried weighting with "1" and "-1", but changing the number will give a different result. "0.5" and "-0.5".

How to download and install flat2 :

flat2 is available from Hugging Face
flat2 is available in Hugging Face

Source: Hugging face

flat2 is available in Hugging Face. The creator seems to be Japanese.

stable-diffusion-webui
stable-diffusion-webui

The method of introduction is the same as other LoRA. After downloading, just throw it into the Lora folder in [stable-diffusion-webui] → [models].

In this article, we reported the results of testing with a 2D model, but flat2 is also highly effective with real models. Please try to easily improve the quality of your illustrations.

I think that negative LoRA will probably be developed at an explosive speed with this as a starting point, so I would like to try various things in the future.
Next Post Previous Post
No Comment
Add Comment
comment url