OCR0017: Creating Font by taking 10 glyphs of each Tibetan characters and ligatures. (4)

[10kalden]

Description: The objective of the task is to create a digital font from the print font of the major publishers such as Derge, Pecing and Shul by taking 10 glyphs of each Tibetan character. The font creation process is previously implemented by taking one random glyph for each Tibetan character.

• The process involves taking the annotated glyphs and converting them into SVG format.
• The SVG's of the glyphs are then parsed and converted into glyphs format suitable for integrating into a TTF file to create the font.
• The font creation process is done by replacing the glyphs in an existing font with our glyph and creating a new font.
• The existing font is taken as Sambhota Unicode or MonlamTB.
• This process was successful in creating a font from the publishers.

Since The Derge Pechas are printed on wooden blocks, there are many variations for the same Tibetan characters and ligature. To include all the variations of the characters and ligatures in the font, 10 glyphs from each character are taken in the font creation process.

• Requires minor research and exploration to display variation in character in the font.

Completion Criteria: Fully completed font in TTF format with most of the variation of the character included.

Font files: https://drive.google.com/drive/folders/1vqTJJ4ThcGckN1OtfUyheJZkYq2ggIOY?usp=drive_link

Implementation Plan:

Sub-task:

• [x] Modify the pipeline to download 10 random glyphs

• [x] Process all the images

• [x] Explore font creation software to study the glyph randomizing mechanism for character variation

• [x] Do RnD for OTF feature to include the variation in the Tibetan glyphs

• [x] Create the font for the desired publication with all the character variations included

• Due to the limitation of cycling through the glyph variant in font created by using existing font, an alternate approach to create font without using existing font is being implemented.

Sub-task:

• [x] Write a script for setting a new font file with all the glyphs
• [x] Define a standard glyph metrics for all the glyphs, there are two approaches for this.
• [x] If the approach is to use the metrics defined in the SVG file name, write a script to convert the value in font units.
• [x] Set the font meta data.
• [ ] Create the font and do further testing

[10kalden]

• Did some exploration of Calligraphr software, a font creation tool using written handwriting. It can create variations of glyphs with different stylistic sets.

https://www.calligraphr.com/en/webapp/app_home/?/

• Created fonts from Derge glyph's png in Calligraphr to understand glyph variation and randomisation. Evaluated the font using FontLab. the fonts created from Calligraphr consist of the OTF feature. where variations of glyphs are stored as SS (stylistic set variant) glyph name uniXXXX.ss01, uniXXXX.ss02 etc.

[10kalden]

According to my research, I found two ways to include the variant glyph in the font.

1. Stylistic Set using the OTF feature Stylistic set, we can create stylistic sets for different print styles. it will group the alternate glyphs. but using this requires manually changing the stylistic set according to need when using a word processor like MS Word. Randomization could be possible through the scripting method. https://typedrawers.com/discussion/1357/how-can-i-randomize-letters-in-a-typeface https://graphicdesign.stackexchange.com/questions/11342/how-does-random-choice-of-alternate-characters-work-in-opentype-fonts

2. Contextual Alternate currently modifying the pipeline to add glyphs as contextual alternates using the OTF feature, I will test with this method to randomize the glyphs. https://forum.glyphsapp.com/t/stylistic-alternates-vs-contextual-alternatives/13478

[10kalden]

• successfully added all the glyph variants as contextual alternates in the font, needs testing

• this image is from FontLab to visualise all the glyphs in the font

• need to write the rules to specify when to substitute glyphs to define Calt feature, need to think about it.

• I tried using this rule, not working

the Calt feature rules are defined like this

``` feature calt {
    # Substitute 'a.alt1' for 'a' when 'a' is followed by 'b'
    sub a' b by a.alt1;
    # Substitute 'a.alt2' for 'a' when 'a' is followed by 'c'
    sub a' c by a.alt2;
    # Add more rules as needed...
} calt;

[10kalden]

• created font with all the variant glyphs, need to write rules for Calt feature to stimulate randomness

[10kalden]

issue#01: added a rule for simulating randomness, but it replaced the already present rule in the existing font for ligature replacement. sol#01:

1. I tried extracting the feature file from the existing font and appending the Calt feature file to it, but it didn't work.
2. currently, tried this method: did some research for ligature replacement rule and wrote a new feature file for ligature replacement. so instead of having one feature file with both Liga and Calt features, wrote a separate feature file. currently fixing some bugs and require testing

rules for ligature replacement

[10kalden]

• created font with 10 glyph variants for each glyph and ligature.
• Added rule for cycling through all the alternate glyphs when typing.
• tested in Libre office, needs more testing. doesn't work in all word processors, it needs a word processor with the OTF feature. link to font

[10kalden]

• successfully made font with contextual alternates, but they still need more diverse rules to be written. they only switch between 2 glyphs.
• testing some alternate rules to get the desired result.

[10kalden]

• The reason for the glyph alternating between 2-3 glyph variant is that because of having 9 variant for each character that comes around 2500 variant in total + 350 original glyphs. so when you write the rule for each permuation. it exceed the limit of substitution.
• There are practical limits based on the size of the LookupList table, which stores the substitution data. Each substitution adds to the size of this table, and if the table becomes too large, it can cause an overflow error. https://adobe-type-tools.github.io/afdko/OpenTypeFeatureFileSpecification.html
• need to find solution for this or either go with 2-3 variant of glyphs.

font link to single gylph and variation glyph: https://drive.google.com/drive/u/0/folders/1vqTJJ4ThcGckN1OtfUyheJZkYq2ggIOY

[ngawangtrinley]

@10kalden What is the reason you are not creating a font from scratch?

[10kalden]

• @ngawangtrinley I considered making the font from scratch. I listed the pros and cons of making a font from scratch or using an existing one. Making a font from scratch is more time-consuming and more complex. whereas using an existing font was less complex and less time-consuming.
• to generate synthetic data, I thought both processes would yield the same result, so I chose to use an existing font.
• I will make it from scratch if there is any issue with using an existing font. I chose to use an existing font as it has more pros than cons for me.

[ngawangtrinley]

It looks like you are running into limitations with existing fonts though, you don't have enough codepoints and it seems like you also have issues with ligatures. Can you give it a try? Now you are quite familiar with font building and you can get help from ChatGPT or Gemini. If you ask AI to give you code one step at a time you'll get good results.

You can try to use the command line of fontforge from python scripts if you don't get much luck with fonttools (by asking AI to write the code to start of course).

A few months ago I did some tests with the help of perplexity.ai and got very good results. We shouldn't be limited by the number of codepoints of old fonts.

Here's a sample chat with Gemini to give your an idea of what I mean: https://gemini.google.com/share/6e511e00b17e.

You can also ask for alternative approaches like https://gemini.google.com/share/9403ff9c1343.

[10kalden]

@ngawangtrinley Yes, I was running into issues with the variants due to the limitations of the lookup table of the existing font.

I will make the font without using an existing font. I have already started with the script. I get help from AI also.

[10kalden]

• wrote a script for making fonts from scratch for only a single variant of each character. need to set glyph metrics
• the issue with character spacing was due to the adjustment of the existing font metrics and using that as the new metrics for the new font.
• wrote a script to convert the metrics we have into font units to set custom metrics for individual glyphs

• ྅_24_6_7.svg this is an example of the file name for SVG, where 24 is the glyph width, 6 is the lsb and 7 is the rsb in pixel

to solve the spacing issue, we need to adjust the advance width advance width = glyph width + lsb + rsb

[10kalden]

@ngawangtrinley hello, I set up a font file using python, but i am running into various issues when setting up some tables by using fonttools. I found an empty font file its called AdobeBlank.ttf(kind of like a templete). I was thinking if we can add all the glyph to this font, set our own metrics. what do you think of this approach. it'll solve the limitation issue. If not, I will try using the Fontforge CLI.

UPDATE: found a way to make font from scratch without using a blank font

[10kalden]

update: got suggestion on font forum to use the method of blank font files, used adobeblank.ttf and created font form scratch.

[10kalden]

got more information about creating a font without using Adobe Blank, using fontbuilder class, tried it this morning and created a font from scratch without taking any blank font. need to test which is better and will take suggestions.

@kaldan007 font for testing https://github.com/OpenPecha/create-font-from-glyph/tree/feature/font-from-scratch/fonts/derge_font

Blocker:

• We don't have glyphs for several characters such as ྲ and ྱ and many more.
• we have around 400 glyphs and many are ligatures, so if we don't have ྱ , we won't be able to type ཀྱ
• so the glyphs won't render when typed.