[Story]CJK fonts, do we have to pay attention to even these things?

Sandoll's New Challenge: IBM Plex Sans Series" 

Development Background of the IBM Plex Sans Series

As the global market expands, there's growing interest in multilingualism, and with this increased interest, the need for multilingual design has also risen. Designing multilingual characters to align with a company's philosophy or direction is crucial to ensure that users encounter a consistent brand image, regardless of the characters they see.

Following this trend, there's been an increase in the design of CJK fonts, which encompass Chinese (C), Japanese (J), and Korean (K) characters, including Hanja, Kana, and Hangul. Designing these characters requires creating thousands to tens of thousands of characters. Since the usage and appearance of characters have evolved differently in each country, many local font foundries often design the characters specific to their country when developing CJK fonts. For example, Chinese characters are designed by Chinese foundries, and Japanese characters by Japanese foundries.

Although local foundries can ensure design expertise due to their deep understanding of the characters, it's challenging to ensure that these designs align well with the overall brand direction, especially for closely related languages like CJK. Therefore, Sandoll has been working with IBM to design the CJK characters for the multilingual font series IBM Plex Sans.The [IBM Plex project] is a proprietary font project that IBM has released under the Open Font License (OFL), with fonts for a total of 8 language groups already published. The Korean version, IBM Plex Sans KR, was released in June 2020, and the Japanese version, IBM Plex Sans JP, in July 2021. The Chinese version is also being developed and directed by Sandoll and is scheduled for future release.


Plex Sans Latin Concept

IBM Plex Sans is based on IBM's historically deep-rooted logotype design. The most critical design keywords for the Plex series are human/machine and natural/engineered, reflecting the philosophy of aiding the advancement of humans and technology together. The Latin version is based on the grotesque style, using angular curves for the counters and natural curves for the letter outlines, creating a distinctive difference in the curvature inside and outside the letters. This feature is especially noticeable in letters such as 'P', 'g', '8'.

Challenges of Matching Latin/CJK

We had to extend the CJK characters based on the already released IBM Plex Sans Latin alphabet, which presented many considerations from the beginning of the project. Since Latin alphabets and CJK characters have different composition, alignment, and developmental methodologies, we needed to discuss how and to what extent we could apply the design concept inherent in the Latin alphabet to the CJK characters.


Challenges Arising from Different Composition Methods

As known, Latin alphabet, Kana, Hangul, and Hanja have different composition methods. For example, while Latin and Kana spell out sounds, Hangul combines consonants and vowels within a single character to produce a sound, and Hanja combines characters with meanings to form another character with a different meaning. This difference in composition means that, unlike Latin and Kana, Hanja and Hangul require designing a wide range of stroke counts within a single character.

To design characters with varying stroke counts to match the Latin alphabet, one must consider "grayscale." "Considering grayscale" means checking and adjusting the balance between the black and white parts of a character. For example, ensuring that no single letter or stroke appears too dark or too light within a sentence. Adjusting the grayscale for CJK fonts to match the Latin alphabet standards can make the characters appear to have a uniform grayscale at certain sizes but look like they have different weights at others. Finding the right balance of grayscale between Latin and CJK characters is essential.

For instance, "Image 1" shows a more uniform grayscale on the right side without any darker letters compared to the left. However, will this uniformity remain when viewed at larger sizes?

 Image 1

"Image 2" is an enlargement of "Image 1". While the characters on the left appear to have the same weight, those on the right seem to have different weights. This variability in grayscale with size means that achieving perfect uniformity at certain sizes is less important than finding an appropriate balance of thickness. 

Image 2

Especially for Japanese, which uses a combination of "Katakana - Hiragana - Kanji" alongside Latin, balancing these characters involves setting the stroke-count-minimal Katakana as the thickest and smallest, and Kanji as the thinnest and largest. This is because characters with fewer strokes appear larger if they are the same size as those with more strokes, and setting them smaller compensates for this. Additionally, characters with fewer strokes are made thicker to prevent them from appearing as if they have holes, which would make the grayscale appear uneven compared to characters with more strokes. 

 Challenges Arising from Visual Centerline Differences

The perspective on the size and positional relationship of languages varies slightly, and each language group has its standard settings for characters, necessitating careful consideration of where and how large to set CJK characters. Latin alphabets are designed based on several lines, such as the Baseline, X-height, Ascender, and Descender, which determine the height of uppercase and lowercase letters.

In contrast, CJK characters, with their diverse stroke counts and combinations, are designed to fit well within a frame based on a "visual centerline." Integrating characters drawn with different standards can result in misaligned positions or sizes. Several strategies must be considered to solve these issues, such as deciding on the size of Latin designs to fit CJK, adjusting CJK sizes to match Latin, and differentiating within CJK characters themselves in terms of size and alignment.

For IBM Plex Sans KR compared to JP, adjustments were made based on local Width and Height settings, and the size of the Latin alphabet was increased by approximately 105% due to it appearing relatively smaller than the specially designed Japanese characters according to specific standards.


Challenges Arising from Element Differences

Another major topic when matching multilingual fonts is design elements. Design elements are crucial for making characters from different scripts appear as part of a single font family, requiring a degree of unification in the elements' imagery across different characters. For example, the impact of curves in the Latin alphabet on the overall font is significant. Fortunately, Korean has circular elements like 'ㅇ, ㅎ', but Chinese does not. How do we incorporate these features into the CJK font?


Chinese characters include elements that, while not circular, allow for the expression of curves, such as 'Dot', 'Hook', 'Downstroke to Left', 'Downstroke to Right', 'Shoulder'. Testing these elements' curvatures to closely match the impression of the Latin alphabet is key. IBM Plex Sans JP reflects the harmonious balance between natural curves and straight sections, and mechanical stroke endings in Kana. IBM Plex Sans KR incorporates the characteristic curvature of IBM Plex Sans in elements like 'ㅇ' and 'ㅎ'.

In reflecting the features of IBM Plex Sans in JP and KR, the most critical consideration was finding the right point where these features, when applied to each script, did not feel "unnatural." Efforts were made to incorporate features in a way that did not feel awkward within the direction of the strokes, ensuring that the Latin alphabet's characteristics did not feel out of place when applied to Chinese and Korean characters.