Wetlook World Forum

Your prompting is going to be dependent on what kind of model you use. If you use for instance a Stable Diffusion 1.5/XL model you will likely have to put your prompts in CLIP format. But if you use something like Flux or one of its fine-tunes, you will use T5 format, which looks a lot more like regular English.

If you're into Flux, check this one out. https://civitai.com/models/141592/pixelwave

and this Lora: https://civitai.com/models/850527/flux-wet-look?modelVersionId=951586

One other thing: you can use tag-style (CLIP-style) prompts with Flux. Sometimes the results are quite good. I used this prompt once to generate a photo that got very positive results:

"asymmetrical composition, full closeup, gritty, film photo, snowy outdoors, dusk moody high-end public modern building, outdoor public bath, steamy soaking pool, blacks, blues, browns, yellows, glass, modern design, guggenheim, rome four seasons, old money, MoMA, tokyo, seoul, high contrast chiaroscuro, weirdo colors, she is sitting in the hot soaking water leaning against the wall of the pool, water up to her stomach, with her arms outstretched to her side, resting on the sides of the pool, she wears a black silk sleeveless boat-neck tee-shirt, knee-length dark grey tartan plaid wool pencil skirt , black pantyhose, Fujifilm XT3, Canon R5, Fujicolor Fujichrome Velvia 100"

produced this: https://imgur.com/a/LTBPAZ4

You'll also want to experiment with all the parameters, like the schedulers and samplers, steps and CFG values. ANd Just wait until the de-distilled checkpoints are available. THat's going to be great.

But in the end, you'll have to train your own model if you want good results. At least that's the only way I've gotten good results!