Golden Knight (Extremely Low CFG 1.10) Part 1
I want to show you the huge difference between two styles using practically the same prompt, models, and LoRAs, simply by changing some of the CFG values. In this image, the CFG is extremely low. CFG is an indicator and a value related to how closely the generation model follows the prompt, as well as how much “creative freedom” it has. The lower the value, the more freedom the model has to interpret the instructions. At some point, when the CFG is extremely low, most of the instructions you put in the prompt become barely noticeable or even completely absent from the final result because the model either ignores them or only loosely follows them. In this case, because the CFG is much lower than the recommended range, the image looks somewhat blurry and diffuse despite having a considerable number of steps. I’ll show you another image using a very similar combination, but with a higher CFG value, and you’ll be able to notice the huge difference between the two images.
Show Parameters
- Size
- 1280 x 992
- Sampling Steps
- 25
- Sampling Method
- Euler a
- CFG Scale
- 1.100000023841858
- Negative
- sword 1,3, blade 1,3, bad quality, worst quality, lowres, jpeg artifacts, bad anatomy, bad hands, multiple views, signature, watermark, censored, ugly, messy

