Hugging Face's logo

Totsukawaii
/
LeagueBase

Text-to-Image
stable-diffusion
Model card Files
xet
Community
Totsukawaii commited on
Commit
b262f67
·
1 Parent(s): 49726cf

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +24 -7
README.md CHANGED
@@ -9,29 +9,46 @@ tags:
9
 
10
  ----
11
 
12
- # Riot Diffusion XL
13
 
14
- Trained on Riot Games Artworks + anime styled images
15
 
16
- ![](https://media.discordapp.net/attachments/1021358999389605950/1138703788883054632/00025-258304652.png)
17
 
18
- ## Notes
 
19
 
20
- - V1.0 is still undertrained
21
- - No noise offset
22
 
 
23
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
  ----
26
 
27
 
 
 
28
  Join my discord server - https://discord.gg/2RBWXFuVV6
29
 
30
  ----
31
 
32
  Credits:
33
  - NAI - base model
34
- - Riot Games - majority of artwork
35
  - Kohya ss sd scripts - trainer
36
 
37
 
 
9
 
10
  ----
11
 
12
+ # lolsplashart, league of legends splash art base
13
 
14
+ High quality model trained for 400 hours on a 4090 on all League Splash Arts up till Snow Moon Skin Line.
15
 
16
+ ![00236-3892470181](https://github.com/Totsukawaii/imgup/assets/45559904/fddb3e51-d998-44a5-a4fd-14bad5b20a00)
17
 
18
+ PROS:
19
+ Diverse outputs/ Not overfitted ✔️
20
 
21
+ NAI base ✔️
 
22
 
23
+ Low Learning rate 3e-7 ✔️
24
 
25
+ High training resolution 2048x1152 ✔️
26
+
27
+ Better than all the overfitted models on civitai ✔️
28
+
29
+ CONS:
30
+ Slightly underfitted
31
+
32
+ Bad anatomy
33
+
34
+ Some outputs appear sunburned
35
+
36
+ Trained with inaccurate tagging - automatic tagging with wd14 tagger, blip
37
+
38
+ Cannot recreate original training images
39
 
40
  ----
41
 
42
 
43
+ *If using this model in your merge please include the recipe in your release.
44
+
45
  Join my discord server - https://discord.gg/2RBWXFuVV6
46
 
47
  ----
48
 
49
  Credits:
50
  - NAI - base model
51
+ - Riot Games - artwork
52
  - Kohya ss sd scripts - trainer
53
 
54