Summary

Can add a little or a lot of painterly style + more interesting compositions to generations. Works on well on 2.5d or Anime style but if you weight it harder it can work on photoreal. Trained on oil painting artist Andrew Esao known for their works that can be stunning or creepy. Simple prompts work better. For fun try "creepy" or "horror" (if thats your thing).

Positive prompt:

Use simple prompts with out much weighting. Terms like surreal, flowery background, still life etc can encourage some very interesting compositions. There is no trigger word however if you would like to increase the effect, add style is oil painting or i.e oil painting of woman

depending on the model

Negatives:

For negatives (on human subjects): Monochrome, 3d render, photo, digital art, signature, infant, kid

For negatives in anime style models: Monochrome, photo, anime, 3d render, lineart, digital art, signature, infant, kid

No need for weights in negatives.

Misc

From the testing that I have done, it seems to be usable across a wide variety of weights, however this depends on the model its used on.

SD 1.5, clip skip 1 was used as the training model (intentionally)

I reccomend Nothing v2.3, Galena REDUX or a well trained flexible model that can do near photoreal and anime.

描述:

Initial Version

训练词语: