Disty0 commited on
Commit
264905e
1 Parent(s): b8ea55d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +121 -0
README.md CHANGED
@@ -3,6 +3,127 @@ pipeline_tag: text-to-image
3
  license: other
4
  license_name: faipl-1.0-sd
5
  license_link: LICENSE
 
 
 
6
  prior:
7
  - Disty0/sotediffusion-v2-prior
8
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  license: other
4
  license_name: faipl-1.0-sd
5
  license_link: LICENSE
6
+ tags:
7
+ - text-to-image
8
+ - diffusers
9
  prior:
10
  - Disty0/sotediffusion-v2-prior
11
  ---
12
+
13
+
14
+ # SoteDiffusion V2
15
+
16
+ An anime diffusion model finetuned on Würstchen V3.
17
+
18
+ # Release Notes
19
+
20
+ - This release is sponsored by <a href="https://fal.ai/grants?rel=sote-diffusion" target="_blank">fal.ai/grants</a>
21
+ - Trained on 12M text & image paris including WD tags and natural language captions for a single epoch on 8xH100 80GB SXM5 GPUs.
22
+ -
23
+
24
+ <style>
25
+ .image {
26
+ float: left;
27
+ margin-left: 10px;
28
+ }
29
+ </style>
30
+
31
+ <table>
32
+ <img class="image" src="https://cdn-uploads.huggingface.co/production/uploads/6456af6195082f722d178522/KJTHqR3otoKoiXxvbudp8.png" width="320">
33
+ <img class="image" src="https://cdn-uploads.huggingface.co/production/uploads/6456af6195082f722d178522/uua4L9aaqJ0LI8gYv4xmC.png" width="320">
34
+ </table>
35
+
36
+
37
+ ## WD Tags:
38
+
39
+ Model is trained with this tag order:
40
+ ```
41
+ aesthetic tags, quality tags, date tags, custom tags, rating tags, character, series, rest of the tags
42
+ ```
43
+
44
+ ### Date:
45
+
46
+ | tag | date |
47
+ |---|---|
48
+ | **newest** | 2022 to 2024 |
49
+ | **recent** | 2019 to 2021 |
50
+ | **mid** | 2015 to 2018 |
51
+ | **early** | 2011 to 2014 |
52
+ | **oldest** | 2005 to 2010 |
53
+
54
+ ### Aesthetic Tags:
55
+ **Model used**: shadowlilac/aesthetic-shadow-v2
56
+
57
+ | score greater than | tag | count |
58
+ |---|---|---|
59
+ | **0.90** | extremely aesthetic | 125.451 |
60
+ | **0.80** | very aesthetic | 887.382 |
61
+ | **0.70** | aesthetic | 1.049.857 |
62
+ | **0.50** | slightly aesthetic | 1.643.091 |
63
+ | **0.40** | not displeasing | 569.543 |
64
+ | **0.30** | not aesthetic | 445.188 |
65
+ | **0.20** | slightly displeasing | 341.424 |
66
+ | **0.10** | displeasing | 237.660 |
67
+ | **rest of them** | very displeasing | 328.712 |
68
+
69
+ ### Quality Tags:
70
+ **Model used**: https://huggingface.co/hakurei/waifu-diffusion-v1-4/blob/main/models/aes-B32-v0.pth
71
+
72
+ | score greater than | tag | count |
73
+ |---|---|---|
74
+ | **0.980** | best quality | 1.270.447 |
75
+ | **0.900** | high quality | 498.244 |
76
+ | **0.750** | great quality | 351.006 |
77
+ | **0.500** | medium quality | 366.448 |
78
+ | **0.250** | normal quality | 368.380 |
79
+ | **0.125** | bad quality | 279.050 |
80
+ | **0.025** | low quality | 538.958 |
81
+ | **rest of them** | worst quality | 1.955.966 |
82
+
83
+ ## Rating Tags:
84
+
85
+ | tag | count |
86
+ |---|---|
87
+ | **general** | 1.416.451 |
88
+ | **sensitive** | 3.447.664 |
89
+ | **nsfw** | 427.459 |
90
+ | **explicit nsfw** | 336.925 |
91
+
92
+ ## Custom Tags:
93
+
94
+ | dataset name | custom tag |
95
+ |---|---|
96
+ | **image boards** | date, |
97
+ | **text** | The text says "text", |
98
+ | **characters** | character, series
99
+ | **pixiv** | art by Display_Name, |
100
+ | **visual novel cg** | Full_VN_Name (short_3_letter_name), visual novel cg, |
101
+ | **anime wallpaper** | date, anime wallpaper, |
102
+
103
+
104
+ ## Limitations and Bias
105
+
106
+ ### Bias
107
+
108
+ - This model is intended for anime illustrations.
109
+ Realistic capabilites are not tested at all.
110
+
111
+ ### Limitations
112
+
113
+ - Can fall back to realistic.
114
+ Add "realistic" tag to the negatives when this happens.
115
+ - Far shot eyes and hands can be bad.
116
+ - Still has a lot of room for more train.
117
+ - Diffusers outputs aren't as good as ComfyUI outputs.
118
+
119
+
120
+ ## License
121
+
122
+ SoteDiffusion models falls under [Fair AI Public License 1.0-SD](https://freedevproject.org/faipl-1.0-sd/) license, which is compatible with Stable Diffusion models’ license. Key points:
123
+
124
+ 1. **Modification Sharing:** If you modify SoteDiffusion models, you must share both your changes and the original license.
125
+ 2. **Source Code Accessibility:** If your modified version is network-accessible, provide a way (like a download link) for others to get the source code. This applies to derived models too.
126
+ 3. **Distribution Terms:** Any distribution must be under this license or another with similar rules.
127
+ 4. **Compliance:** Non-compliance must be fixed within 30 days to avoid license termination, emphasizing transparency and adherence to open-source values.
128
+
129
+ **Notes**: Anything not covered by Fair AI license is inherited from Stability AI Non-Commercial license which is named as LICENSE_INHERIT.