Update README.md
Browse files
README.md
CHANGED
|
@@ -114,20 +114,13 @@ The workflow is available in this repository. The comparison images are also the
|
|
| 114 |
|
| 115 |
**Universal Sampling Parameters:**
|
| 116 |
|
| 117 |
-
Resolution: 1344x768
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
Scheduler: exponential
|
| 126 |
-
|
| 127 |
-
Denoise: 1
|
| 128 |
-
|
| 129 |
-
CLIP Scaling: 4
|
| 130 |
-
|
| 131 |
Negative Prompt: text, watermark, logo, blurry, out of focus
|
| 132 |
|
| 133 |
These parameters gave me the best results - according to my preference, but also for the sake of comparison. They do lean a bit more to the tame side of things to keep random variance down, so I'd recommend experimenting and pushing the boundaries of these parameters if you decide to play around with these models yoruself. I've gotten some good results with CFG set to 1.5, and making use of the Dynamic Thresholding extension (available for both Stable Diffusion and ComfyUI) to boost some of the CFG effects significantly.
|
|
|
|
| 114 |
|
| 115 |
**Universal Sampling Parameters:**
|
| 116 |
|
| 117 |
+
Resolution: 1344x768
|
| 118 |
+
Steps: 45
|
| 119 |
+
CFG: 4
|
| 120 |
+
Sampler: dpmpp_3m_sde_gpu
|
| 121 |
+
Scheduler: exponential
|
| 122 |
+
Denoise: 1
|
| 123 |
+
CLIP Scaling: 4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 124 |
Negative Prompt: text, watermark, logo, blurry, out of focus
|
| 125 |
|
| 126 |
These parameters gave me the best results - according to my preference, but also for the sake of comparison. They do lean a bit more to the tame side of things to keep random variance down, so I'd recommend experimenting and pushing the boundaries of these parameters if you decide to play around with these models yoruself. I've gotten some good results with CFG set to 1.5, and making use of the Dynamic Thresholding extension (available for both Stable Diffusion and ComfyUI) to boost some of the CFG effects significantly.
|