niji_ancient_style_xl

Standing on the shoulders of giants.....

万象熔炉 | Anything XL

Some adjustments were made based on the foundation of the great artist, and the data was still done by niji pictures sharing their own.

This XL model, you can achieve a good ancient style effect with simple SD1.5 type keywords like 1girl, solo, hanfu,:

Alternatively, you can achieve an excellent ancient style effect by describing in natural language like the most beautiful woman in China, dressed in light blue and cyan hanfu with hair up, standing on the grassland by Lake Asal. She is wearing exquisite makeup, with a long flowing skirt, a smiling expression, soft lighting, surrounded by natural scenery and bright colors,:

This makes it easier for everyone to use.

This time, using the X/Y/Z plot method to show everyone the influence of sampling and steps on the output graph:

When step = around 40, the sampling method doesn't seem to have a big impact, so everyone can just follow their own habits.

Also, because of the outstanding ability of 万象熔炉 | Anything XL, even when the ancient style is not stimulated, it still performs very well. The following are the differences between this model and 万象熔炉 | Anything XL under different prompts:

Above is the万象, below is this model, Figures 1, 2, 5 are prompted with the hanfu keyword, while 3, 4 are normal anime pictures

So ultimately it's still the fact that the original model is too powerful, thank you.

This model is mainly for generating images for tweets, producing some images that match novels.

Since the images are all self-generated and not carefully selected, leading to some imperfections, it is still recommended to enable ADetailer for touch-ups. As for how to improve in the future, hmm, still learning and working on it, you can also try using the lora of xl for retouching hands...

For example: Hand Repair ClearHands

Regarding dimensions, both horizontal and vertical are possible: 1024*768, 768*1024, and other common sizes

Welcome back for more images~