wlop / README.md
SirVeggie's picture
Update README.md
6bdffe4
metadata
license: creativeml-openrail-m
tags:
  - art

WLOP stable diffusion model

Original artist: WLOP
Patreon: https://www.patreon.com/wlop/posts

Basic explanation

Token and Class words are what guide the AI to produce images similar to the trained style/object/character. Include any mix of these words in the prompt to produce verying results, or exclude them to have a less pronounced effect. There is usually at least a slight stylistic effect even without the words, but it is recommended to include at least one. Adding token word/phrase class word/phrase at the start of the prompt in that order produces results most similar to the trained concept, but they can be included elsewhere as well. Some models produce better results when not including all token/class words.

For model merging I recommend using the wlop.ckpt or wlop-any model.

Model: AbyssalWlop - (current best version)

The model works without a keyword, but you can affect the style with the keywords m_wlop illustration style, which are used by the merged models.

The model works best at clip skip 2 and 3.

Mix using AbyssOrangeMix2_nsfw, wlop and wlop-any models to create a stable and accurate wlop style. The recipe itself is quite simple.

orange-wlop    = AbyssOrangeMix2_nsfw + (wlop-any - anything) @1.0
orange-wlop2   = AbyssOrangeMix2_nsfw + (wlop     - wd1.3)    @1.0
AbyssalWlop    = orange-wlop          + orange-wlop2          @0.5

Image comparisons between models, more models located under the image grids:

grid1 grid2

Model: wlop-any

Has the most consistent wlop style, but difficult to get good results

token: m_wlop
class: illustration style
base: anything v3
images: 120
steps: 12000

Model: wlop-anymix

Custom berry mix using wlop-any as last step. Great quality if prompted correctly, but loses wlop style. Is influenced by the style though.

Model: wlop

This version is highly overfit, and not suitable for standalone use. Merge with another model to use.

token: m_wlop
class: illustration style
base: waifu diffusion 1.3-full
images: 160
steps: 16000

Model: wlopmix

Custom berry mix using wlop as last step. Pretty similar to wlop-anymix, though there are some flavor differences.

Model: wlop_e5

Old wlop model, I guess it works ok. Decent wlop style reproduction if you can get good quality out of it.

token: m_concept
class: 1girl
base: waifu diffusion 1.3-e5

License

This embedding is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:

  1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content
  2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
  3. You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here