FoodDesert commited on
Commit
060acf3
·
1 Parent(s): 2b88ee9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +13 -11
README.md CHANGED
@@ -1,7 +1,7 @@
1
 
2
  # boring_e621
3
 
4
- This embedding attempts to capture what it means for an image to be uninteresting. It was trained as a negative embedding for [ye18](https://huggingface.co/Doubleyobro/yiffy-e18) using e621 style tags as prompts during training.
5
  If you're using the [Automatic1111 Stable Diffusion WebUI](https://github.com/AUTOMATIC1111/stable-diffusion-webui), place the .pt file in
6
  stable-diffusion-webui\embeddings and add "by boring_e621" to your negative prompt for more interesting outputs.
7
  <br>
@@ -9,7 +9,7 @@ stable-diffusion-webui\embeddings and add "by boring_e621" to your negative prom
9
  ## Model Description
10
 
11
 
12
- The motivation for this is that negative embeddings like [Bad Prompt](https://huggingface.co/datasets/Nerfgun3/bad_prompt),
13
  whose training is described [here](https://www.reddit.com/r/StableDiffusion/comments/yy2i5a/i_created_a_negative_embedding_textual_inversion/)
14
  depend on manually curated lists of tags describing features people do not want their images to have, such as "deformed hands". Some problems with this approach are:
15
  * Manually compiled lists will inevitably be incomplete.
@@ -18,21 +18,21 @@ depend on manually curated lists of tags describing features people do not want
18
  <br>
19
 
20
  To address these problems, boring_e621 employs textual inversion directly on a set of 100 of the most boring images on the furry image site e621.net.
21
- These images were automatically collected by querying e621 for images that no user has ever Favorited or Up Voted.
22
  Almost all images on the site have at least one up vote or favorite, so the images that no one favorited are some of the very most uninspiring images on the website.
23
  Thus, you should be able to use the resulting embedding as a negative prompt to make your generated images Not Boring.
24
  <br>
25
 
26
  # Bias, Risks, and Limitations
27
  * Using this as a negative embedding often sacrifices some fidelity to the prompt. For example, characters in the image may disappear or change eye/skin/fur color.
28
- * Using this as a negative embedding may introduce sexual content into images that would not otherwise have it. I suspect this is because people do not find sexual content boring, and thus it may be absent in the embedding's training data.
29
  * Unlike other negative embeddings, this is not intended to fix problems like extra limbs or deformed hands. It can be used alongside other negative embeddings to fix deformities.
30
  <br>
31
 
32
  # Evaluation
33
 
34
 
35
- I extracted the tags from three e621 images and used them to construct prompts in the style used by [ye18](https://huggingface.co/Doubleyobro/yiffy-e18).
36
  * one prompt was constructed from an image with a high number of favorites.
37
  * one prompt was constructed from an image with a moderate number of favorites.
38
  * one prompt was constructed from an image with 0 favorites.
@@ -45,7 +45,7 @@ I then generated test images from each of these prompts, each time using a diffe
45
  * [boring_e621](this)
46
  <br>
47
 
48
- Finally, I qualitatively evaluated the attractiveness and interestingness of the resulting images, though I will let you draw your own conclusions from the output.
49
  <br>
50
 
51
  ## Results
@@ -54,9 +54,11 @@ Finally, I qualitatively evaluated the attractiveness and interestingness of the
54
 
55
 
56
  ## Other Models
57
- boring_e621 has been reported to work well with SD 1.4 or 1.5 furry models such as:
58
- https://civitai.com/models/12979/lawlass-yiffymix-20-furry-model
59
- https://civitai.com/models/4698/lawlass-yiff-mix
60
- https://civitai.com/models/15503/kavka-mix
61
- https://huggingface.co/Doubleyobro/yiffy-e18
 
 
62
 
 
1
 
2
  # boring_e621
3
 
4
+ This embedding attempts to capture what it means for an image to be uninteresting. It was trained as a negative embedding using e621 style tags as prompts during training.
5
  If you're using the [Automatic1111 Stable Diffusion WebUI](https://github.com/AUTOMATIC1111/stable-diffusion-webui), place the .pt file in
6
  stable-diffusion-webui\embeddings and add "by boring_e621" to your negative prompt for more interesting outputs.
7
  <br>
 
9
  ## Model Description
10
 
11
 
12
+ The motivation for boring_e621 is that negative embeddings like [Bad Prompt](https://huggingface.co/datasets/Nerfgun3/bad_prompt),
13
  whose training is described [here](https://www.reddit.com/r/StableDiffusion/comments/yy2i5a/i_created_a_negative_embedding_textual_inversion/)
14
  depend on manually curated lists of tags describing features people do not want their images to have, such as "deformed hands". Some problems with this approach are:
15
  * Manually compiled lists will inevitably be incomplete.
 
18
  <br>
19
 
20
  To address these problems, boring_e621 employs textual inversion directly on a set of 100 of the most boring images on the furry image site e621.net.
21
+ These images were automatically collected by querying e621 for images that no user has ever Favorited or Up-Voted.
22
  Almost all images on the site have at least one up vote or favorite, so the images that no one favorited are some of the very most uninspiring images on the website.
23
  Thus, you should be able to use the resulting embedding as a negative prompt to make your generated images Not Boring.
24
  <br>
25
 
26
  # Bias, Risks, and Limitations
27
  * Using this as a negative embedding often sacrifices some fidelity to the prompt. For example, characters in the image may disappear or change eye/skin/fur color.
28
+ * Using this as a negative embedding may introduce unexpected or undesired content into the image to make it look less boring.
29
  * Unlike other negative embeddings, this is not intended to fix problems like extra limbs or deformed hands. It can be used alongside other negative embeddings to fix deformities.
30
  <br>
31
 
32
  # Evaluation
33
 
34
 
35
+ I extracted the tags from three e621 images and used them to construct a set of test prompts.
36
  * one prompt was constructed from an image with a high number of favorites.
37
  * one prompt was constructed from an image with a moderate number of favorites.
38
  * one prompt was constructed from an image with 0 favorites.
 
45
  * [boring_e621](this)
46
  <br>
47
 
48
+ Finally, I qualitatively evaluated the attractiveness and interestingness of the resulting images, though I will let you draw your own conclusions from the output below.
49
  <br>
50
 
51
  ## Results
 
54
 
55
 
56
  ## Other Models
57
+
58
+ Boring_e621 has been reported to work well with SD 1.4 or 1.5 furry models such as:
59
+ * https://civitai.com/models/12979/lawlass-yiffymix-20-furry-model
60
+ * https://civitai.com/models/4698/lawlass-yiff-mix
61
+ * https://civitai.com/models/15503/kavka-mix
62
+ * https://civitai.com/models/17649/bb95-furry-mix
63
+ * https://huggingface.co/Doubleyobro/yiffy-e18
64