Here is the feature description of <Youkengi delicate anime v1.0>:

The characteristics of this large model are: beautiful art style, more details and extremely strong text control ability (which is usually considered as a unique performance of sdxl).

In terms of text control ability, this model is slightly stronger than Youkengi anything anime, but with different focuses.

Replaced the pruned version and merged the vae model.

1. Model Features:

1. Beautiful art style, more delicate details, as shown in the examples.

2. Extremely strong text control ability, manifested in a very strong understanding of words. For example:

(1) Almost always produces full-body images with the prompt of fullbody (even if it's a horizontal image), and with the prompt of back to the viewer, back view images are almost guaranteed (the orientation of the image can almost be ignored, but in cases that do not follow the regular imaging rules, distortion may increase).

(2) Under this model, the effects of prompts alone can almost match some lora effects, do not be surprised to see a tall image in the examples, as it was produced directly with just a prompt. I am personally very surprised as well!

(3) If your word descriptions are precise and stable, with the extremely strong text-to-image control ability, even when switching seeds, the images produced by this model may be very similar in composition. In these cases, you can try words like dynamic pose, girl's pose, which have a certain generalization, to bring out the practical effects of these words and increase composition variations.

(4) You can easily integrate the elements you want into the character with just one word, elements like ears, tails, wings are all present, as illustrated in the xyz example images.

2. How to Use/Generation Suggestions:

1. No specific recommended sampling method: not very picky, the examples were created using dpm++ 2m sde karras;

2. Recommended to use HD restoration redraw amplitude 0.4; based on 25 steps, with 10 steps for HD redraw.

3. If you want to enhance light and watercolor styles: the recommended magnification algorithm is latent, but it can make the image blurry, beginners should use it with caution:

Also, increasing my two lora effects will be more noticeable: using gloss_tweaker:-1, edge_weaken_strengthen:1 as shown in the swimsuit example.

4. If you want to portray a delicate art style: the recommended magnification algorithm is 4x ultrasharp:

5. More versatile would be r-esrgan 4x+, but the model's ********* is strong, so there's not much to worry about.

If you are satisfied, please provide feedback and more images. Additionally, following and bookmarking can keep you updated on future version updates and new models.