Composed Character with StyleGAN
Graphic
Graphic
In graphic design, composed type is a way to integrate more than one character to make a new character rendering to express the meaning of a word. In common ways, designers would design the composed type based on the similar components of different characters and arrange them in a certain way to make a new character. In this project, I tend to take advantage of the algorithm logic instead of graphic method to recreate the composed character.
Research
1. Traditional Method
The design logic of the traditional composed character is formed based on similar components of different characters, such as radicals in Chinese characters or strokes, and then rearrange and combine all the characters. This preserves both readability and aesthetics.
![]()
The design logic of the traditional composed character is formed based on similar components of different characters, such as radicals in Chinese characters or strokes, and then rearrange and combine all the characters. This preserves both readability and aesthetics.
2. Modern Method
A modern example of a composed character is the work of Mei Shuzhi, which is more straightforward than the traditional method. The approach is to split the original two characters and then joint them with similar strokes, using negative shapes and optical illusions.
![]()
A modern example of a composed character is the work of Mei Shuzhi, which is more straightforward than the traditional method. The approach is to split the original two characters and then joint them with similar strokes, using negative shapes and optical illusions.
3. Inverted Method
Inverted combining is reversing two words or characters and combining them into one character or word. Different words could be seen when it is viewed from two directions. Two combined words often have related or opposite meanings, creating a "two sides of the same coin" viewing experience.![]()
Inverted combining is reversing two words or characters and combining them into one character or word. Different words could be seen when it is viewed from two directions. Two combined words often have related or opposite meanings, creating a "two sides of the same coin" viewing experience.
Build Dataset
Based on the three different ways of designing the combined characters, I chose to use two different ways to compose the two types of datasets.
The text content chosen is "什" and "么" (means “what”) because they are simple and can be observed more clearly at the beginning.
StyleGAN Machine Learning Results
![]()
Reflection & Next to do
The essential difference between machine learning algorithms and designers is the semantic understanding of visual elements. While humans design based on strokes, machines generate images based on correlations between pixel points.
The problem of the current solution is the low amount of data, the next step is to try more complex and more data to improve the design.
Future possibilities could be multi-character design, and multi-lingual design, such as the design of combined Chinese and English characters.
The problem of the current solution is the low amount of data, the next step is to try more complex and more data to improve the design.
Future possibilities could be multi-character design, and multi-lingual design, such as the design of combined Chinese and English characters.