Evel/

VividWatercolors

Introducing my new Vivid Watercolors dreambooth model.

The model is trained with beautiful, artist-agnostic watercolor images using the midjourney method.

The token is "wtrcolor style"

It can be challenging to use, but with the right prompts, but it can create stunning artwork.

See an example prompt that I use in tests:

wtrcolor style, Digital art of (subject), official art, frontal, smiling, masterpiece, Beautiful, ((watercolor)), face paint, paint splatter, intricate details. Highly detailed, detailed eyes, [dripping:0.5], Trending on artstation, by [artist]

Using "watercolor" in the pronpt is necessary to get a good watercolor texture, try words like face (paint, paint splatter, dripping).

For a negative prompt I use this one:

(bad_prompt:0.8), ((((ugly)))), (((duplicate))), ((morbid)), ((mutilated)), [out of frame], extra fingers, mutated hands, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), ((ugly)), blurry, ((bad anatomy)), (((bad proportions))), ((extra limbs)), cloned face, (((disfigured))), (((dead eyes))), (((out of frame))), ugly, extra limbs, (bad anatomy), gross proportions, (malformed limbs), ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), mutated hands, (fused fingers), (too many fingers), (((long neck))), blur, (((watermarked)), ((out of focus)), (((low contrast))), (((zoomed in))), (((crossed eyes))), (((disfigured)), ((bad art)), (weird colors), (((oversaturated art))), multiple persons, multiple faces, (vector), (vector-art), (((high contrast)))

Here's some txt2img exemples:

Here an img2img exemple:

In img2img you may need to increase the prompt like: (((wtrcolor style)))

You can play with the settings, is easier to get good results with the right prompt: For me, the sweet spot is around 30 steps, euler a, cfg 8-9. (Clip skip 2 kinda lead to softer results)

See the tests here: https://imgur.com/a/ghVhVhy