Contemporary Mujia

The work absorbs the elegant depiction of external form and contour lines from Japanese woodcuts, the magnificent colors and geometric decorative effects of Byzantine art, and the intricate and sensual portrayal of Baroque and Rococo art. It creates a character image known as the "Mucha style" using sensual decorative lines, simple contour lines, and vibrant watercolor effects.

------------------------------------------------------------------------------

Model Features

-- Distinct style, stable triggering words, strong decorative effects

-- No need for high-definition restoration

-- Compatible with various large models

-- Strong generalization ability, achieving the extreme of a stylized style, with the content mostly depending on the capacity limit of the large models

-- Through incomplete testing, the content that the large model can produce can all be transformed into the "Mucha style" combined with this lora

-- Retro color tones, vivid colors, high brightness, able to effectively reproduce the color capabilities of large models

-- Apart from human error, the rate of image breakdown is very low.

-- Required triggering words: art nouveau, stylized, Alphonse Mucha

-- Reinforced decorative border words: intricate border, decorative border

-----------------------------------------------------------------------------------------------

Basic Usage Instructions: Due to the different characteristics of each large model, it is recommended to have a weight between 0.6-1.2. Besides weightings for style triggers, additional weights for triggering words can also be added, for example (Alphonse Mucha: 1.2) to enhance the style.

-----------------------------------------------------------------------------------------------

In order to restore the ultimate Mucha style, the decorative painting in the model has been maximized, selecting around a hundred high DPI high-definition works and manually color-correcting them, adjusting the color levels, brightness, and saturation of each image, striving to enhance the color rendering of the image without losing its original features.

The model after image color correction has significantly improved the brightness and color rendering of the image compared to direct use of the original materials for training.

Above are the model training comparisons before and after color correction under the same seed, parameters, iterations, and large model (training parameters are also the same)

The model also tested the compatibility effects with the large models commonly used by the author, and most mainstream models can guarantee the image output effect while also reproducing the "Mucha style," achieving a qualification standard for a well-stylized image.

Welcome everyone to try it out more and provide more feedback. Wishing everyone a pleasant experience with creations.