YACAM版本SR⭐33614 (ID: 397419)

YACAM is just an abbreviation for "Yet Another Cartoon/Anime Merge"

There have already been so many merges here and the key idea of this one is using "broken", "over trained", "strange" models as a part. My operations with such models are usually directed to convert their drawbacks to their features. Unlike tons of other (really good) typical models my ones might be less consistent, but likely they can compose a bit more intriguing results.

The numbers in "version" mean just their order, not how they are "better/worse". Please, read their own descriptions.

Pictures

Watching my directly posted pictures here you should expect easy readable prompts without LORAs, embeddings or inpaintings because I'd like to show you the models as-is. But I like playing with adding noise, samplers, steps, upscalers and prompt editing like [::] or [||]. Even two similar pictures nearby may have some difference in the workflow.

Usually I run lots of seed to find a decent picture without a typical finger salad.

What to do next:

1. Check their descriptions and take a look at the images attached

2. Download the chosen model and challenge your prompts

3. Share the results

Keep in mind that all creators need your reactions to know what they do right or wrong. We don't have a "dislike" button, but I consider that the "crying" smile does the same.

Version details

SR — "Semi Realistic". In different versions they can pretend to be a photo-like model, but the main purpose is working with artistic styles. As for me — "true photo realistic" models are boring. I'm ready to trade a bit of "natural skin texture" for "glowing eyes" etc.

CR — "Cartoon". Not only anime style, but any cartoon style at all.

The problem

The author doesn't know what to draw other than nice breasts.

描述:

I'd say that it's my tribute to the creator of "fantasticmix" . I really like the original model, especially v7. It has a very successful combination of semi-realism, anime-like artistic features and an easy recognizable woman face as its signature.

Obviously, as an any LORA-integrated model it has some shortcomings. So my favorite v7 was taken as the base and here I've tried to expand the variety for faces, revive common artistic styles and bring some new content but while keeping the iconic features of the model intact.

Almost all my pictures here were made using Restart sampler and CFG:4, also some images turned out very good even with CFG:2

The faces here can be easily changed using faces mixing [famous_person_1|famous_person_2] or reverted to the original with something like "kpop idol looking" or "ulzzang" in your prompts. Hopefully my prompts can reveal more. Finger are still the most dangerous object and they will hurt you, I believe.

I found useful for negative here the next:

(bad or low quality), low resolution

Try to not overload your prompts.

训练词语: