OCR0046: Adding ligature replacement for stack with Tibetan vowels glyphs and fine tune the font.

10kalden commented 1 month ago

Description:

We have Derge font with single variant of each glyphs, in the font we have ེ ོ ི as glyphs and these glyphs don't have a ligature replacement for their respective ligature. They are being used as individual glyphs.
This is creating an issue where in the font, the glyphs for ེ ོ ི are not perfecting aligned with the tibetan consonants and also there is no variation in these glyphs.

Implementation:

in order to solve this we have to replace the stacks with ེ ོ ི, with their respective ligatures. so that the vowel glyphs will be aligned with their consonants and we could have different variation of vowel glyphs.

in this image the glyph for ོ is not properly aligned with its consonants.

Subtask:

[x] write a script to extract all the ligature with the vowels glyphs
[x] Check if the ligature are present or scan the OPF E-text if they are not present
[x] Fix the placement of various glyphs to match as close as that of the pecha text.
[x] Fine-tune the font when all the stacks and glyphs have been replaced.

Completion Criteria: To obtain a font with ligature replacement for stacks with the Tibetan vowel glyphs.

kaldan007 commented 1 month ago

@ta4tsering kindly review this card as per discussion

10kalden commented 1 month ago

Img.1: Pecha Image of Derge Kangyur

Img.2: Text created from E-text using generated Derge Font

10kalden commented 1 month ago

Img.1: Pecha Image of Derge Kangyur

img.2 Fixed vowel displacement for derge font

10kalden commented 1 month ago

Tibetan Scripts

There are 30 tibetan consonants and 4 vowel marks
Tibetan consonants can be written as consonants or in other forms such as Superscript or Subscripts
When constants are written as superscripts or subscripts, or with vowel letters, there are known as Ligatures
The image is a basic representation of ligatures structure.

10kalden commented 1 month ago

Ligatures

-When creating fonts, we have ligature replacement. when certain letters are typed in sequence according to their unicode value, their respective ligature will be replaced. -In font creation, the superscript vowels ( ི ོ ེ) are independent of the glyphs, meaning they don't require ligature replacement as they will sit on top of whichever glyphs it is typed with. On the other hand the subscript vowel ( ུ ) is heavily dependent on the glyphs and require ligature replacement. -So, If we were to introduce ligature replacement for the superscript vowel, it would require us to crop a whole lot of extra glyphs with all the combination of the above structure.

Tibetan Glyph Count with vowel with 3 superscripts vowels

Consonants 30 consonants: 30 x 3 = 90 glyphs (all the glyphs for this are already present)

Superscripts ར (12 letters): 12 x 3 = 36 ལ (10 letters): 10 x 3 = 30 ས (11 letters): 11 x 3 = 33 Total superscripts glyphs = 99 glyphs

Subscripts ཡ (7 letters): 7 x 3 = 21 ར (13 letters): 13 x 3 = 39 ལ (6 letters): 6 x 3 = 18 ཝ (16 letters): 16 x 3 = 48 Total Subscripts glyphs = 126 glyphs

Combination of super scripts and subscripts རྐྱ་ རྒྱ་ རྨྱ་ སྐྱ་ སྒྱ་ སྤྱ་ སྦྱ་ སྨྱ་ སྐྲ་ སྒྲ་ སྤྲ་ སྦྲ་ སྨྲ་ སྣྲ་ གྲྭ་ དྲྭ་ ཕྱྭ་ རྒྭ་ རྩྭ. 19 glyphs x 3 = 57 glyphs

glyphs required for superscript vowels = 372 glyphs we already have 90 glyphs Total Glyphs required = 282 glyphs

These are the glyph requirement for basic ligatures found most common in the Pechas, but there are some ligature which does not follow the basic ligature structure convention in the pecha and thus the data for those are uncertain.

ngawangtrinley commented 1 month ago

I just tried and don't get the vowels at the right place.

Can you add versioning tags to the font names? Both the file name and inside the font, this way we can be sure we don't have miscommunication.

10kalden commented 1 month ago

@ngawangtrinley Ok, will do that

10kalden commented 1 month ago

@ngawangtrinley can you try this https://github.com/OpenPecha/create-font-from-glyph/tree/feature/font-from-scratch/fonts/derge_font

ta4tsering commented 1 month ago

@10kalden if this ttf is the one that you send me then, it is the same one that @ngawangtrinley has tested.

10kalden commented 1 month ago

The font requires several adjustments as discussed yesterday

[x] reduce the spacing between lines to match that of Pecha
[x] fix vowel placements
[x] baseline for characters ཙ ཚ ཛ
[x] check glyphs which shows unnecessary strokes due annotation
[x] fixed glyphs which are not replaced by ligatures
[x] fix the placement of །
[x] fix the placement of tsek
[ ] some glyphs are missing, find them in the opf
[x] fix the space between sentences

10kalden commented 1 month ago

The generated font glyphs strokes looks thicker, that is due to the vectorization of png images, converting png to svg can help keep the font sharp at any size but it will also smooth out the pecha images glyphs which will cause the fonts to look visuallly different from the source text.

Any thoughts on this @ngawangtrinley

10kalden commented 1 month ago

there are some glyphs which has noise due to incorrect annotation, for that, there is a approach to clean the glyphs using https://github.com/orgs/OpenPecha/projects/56/views/7?pane=issue&itemId=72223820

10kalden commented 1 month ago

as per @ngawangtrinley suggestion, resized the the png 10 times, and the resultant svg show less smoothing of glyphs