Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
|
5 |
+
# EclecticEuphoria_Flux_Unet_Schnell_models
|
6 |
+
|
7 |
+
Not of a huge fan of Schnell on first use so I took up a challenge to remove what I didn't like of Schnell. First goal was to remove the hideous Greenish|Yellowish look and the random daark splotches all over. I used lora and thanks to the negative field was able to remove most of those issues and get a more natural looking image. So I am calling this version a win. The standard model produces much better images out of the gate vs the ggufs. A refiner is definately needed in both cases, the ggufs are complete rubish looking straight out of the sampler, but it's a good thing as the noise gives the refiner something good to work with. So find a favorite refiner model and plug it in for best results. Images are created in 8 steps. I only tested the 8_0K gguf a couple times, not sure how well the K6 and K4 models do. I do like the artsy style of schnell tho, it can really produce some unique beatuiful pictres in a hurry.
|
8 |
+
|
9 |
+
## Standard Workflow
|
10 |
+
|
11 |
+
![](Flux_Unet_Schnell_WF.png)
|
12 |
+
|
13 |
+
## GGUF Workflow
|
14 |
+
![](Flux_Schnell_GGUF_WF.png)
|
15 |
+
|
16 |
+
## Standard Comparison sampler vs refine
|
17 |
+
|
18 |
+
![](sampler_vs_refined.png)
|
19 |
+
|
20 |
+
|
21 |
+
|
22 |
+
## All the upper images were created with this model
|
23 |
+
[Civitai Gallery of Schnell Images from this model](https://civitai.com/collections/1216945)
|