Low CFG 2025

This is the latest series of explorations of Low CFG images with Stable Diffusion.

"CFG" stands for "Classifier-Free Guidance". It is the part of the Stable Diffusion algorithm that controls how much the model tries to fit the output image to the prompt vs its own process or "personality".

Lately I have been experimenting with the idea of Low and No CFG images, which results in some very interesting and unique results. Recognizing that a generative image model like Stable Diffusion has been trained on a large dataset of images from the internet (the LAION-5B dataset of 5.85 billion labeled images), then these image outputs are like proto-images or summations of lots of concepts together, possibly. It is exploring the nature of the generative image model vs trying to replicate the inputs that were used to train the model as is more typical in AI art.

1
1

2
2

3
3

4
4

5
5

6
6

7
7

8
8

9
9

10
10

11
11

12
12

13
13

14
14

15
15

16
16

17
17

18
18

19
19

20
20

21
21

22
22

23
23

24
24

25
25

26
26

27
27

28
28

29
29

30
30

31
31

32
32

33
33

34
34

35
35