Auto / Manual Hinting
Posted: Tue Apr 12, 2005 5:46 pm
About Sorting and Hinting/Grayscale
First, Font Creator is an excellent product. A steep learning curve with powerful features is much better than a dumb learning curve with weak features.
I have just finished a trial run to make an alphanumeric plus connection point font for parsable diagrammatic notations. I partly completed the alphanumeric glyphs and completed the connection point glyphs. I added the connection point glyphs to Lucida (Bright, Sans and Console) to give me something for software development, but I wish to buy Font Creator and complete the project later in the year with auto hinting.
I tested all my glyphs using gray scaling with Windows font smoothing turned on. I estimate that the on screen legibility for 6 pt size was equivalent to the 5 th from bottom row of an opticians eye chart (third row is 20/20 vision and the stokes and gaps subtend 1 minute of arc). The printed results were better.
Auto / Manual Hinting
First my experience of a comparable process is Xilinx’s gate arrays. When first introduced you could either manually or auto place and route gates and registers. About 1990 they introduced incremental place and route with a guide file - a major improvement.
For Font Creator you could flag each glyph as: H hinted, P partially hinted, N not hinted and D don’t hint. A user could then run auto-hint on a per-glyph or all-glyphs basis and test the results. They could then partially un-hint a glyph, repeat the auto-hint on the glyph, and if still not satisfied they could manually hint. The auto-hint algorithm would preserve existing hints as a guide and complete the task.
Why should repeating the auto hint produce different results on a second pass? Well Xilinx incremental place did. I am guessing but a hinting algorithm’s strategy might be to select and order on-curve points and then choose the bit mapped position for them (call that the gridding part) based on a recommended choice preference (left = right – width, right = left + width, width = right – left) (call that the hinting part). Now manually re-hinting only one of the un-hinted points will cause the rest to be different, and even if no un-hinted points were re-hinted the partial hints may change the algorithm’s choice a hinting preferences.
This approach will save a serious amount of work and preserve the designer’s ability to refine the details.
Sorting
When you create a new file you get characters with a background that shows the shape in I guess locale class order. However sorting glyphs in code point order is valuable when manually checking with other documentation e.g. Unicode.org’s unibook.exe.
John Cork
First, Font Creator is an excellent product. A steep learning curve with powerful features is much better than a dumb learning curve with weak features.
I have just finished a trial run to make an alphanumeric plus connection point font for parsable diagrammatic notations. I partly completed the alphanumeric glyphs and completed the connection point glyphs. I added the connection point glyphs to Lucida (Bright, Sans and Console) to give me something for software development, but I wish to buy Font Creator and complete the project later in the year with auto hinting.
I tested all my glyphs using gray scaling with Windows font smoothing turned on. I estimate that the on screen legibility for 6 pt size was equivalent to the 5 th from bottom row of an opticians eye chart (third row is 20/20 vision and the stokes and gaps subtend 1 minute of arc). The printed results were better.
Auto / Manual Hinting
First my experience of a comparable process is Xilinx’s gate arrays. When first introduced you could either manually or auto place and route gates and registers. About 1990 they introduced incremental place and route with a guide file - a major improvement.
For Font Creator you could flag each glyph as: H hinted, P partially hinted, N not hinted and D don’t hint. A user could then run auto-hint on a per-glyph or all-glyphs basis and test the results. They could then partially un-hint a glyph, repeat the auto-hint on the glyph, and if still not satisfied they could manually hint. The auto-hint algorithm would preserve existing hints as a guide and complete the task.
Why should repeating the auto hint produce different results on a second pass? Well Xilinx incremental place did. I am guessing but a hinting algorithm’s strategy might be to select and order on-curve points and then choose the bit mapped position for them (call that the gridding part) based on a recommended choice preference (left = right – width, right = left + width, width = right – left) (call that the hinting part). Now manually re-hinting only one of the un-hinted points will cause the rest to be different, and even if no un-hinted points were re-hinted the partial hints may change the algorithm’s choice a hinting preferences.
This approach will save a serious amount of work and preserve the designer’s ability to refine the details.
Sorting
When you create a new file you get characters with a background that shows the shape in I guess locale class order. However sorting glyphs in code point order is valuable when manually checking with other documentation e.g. Unicode.org’s unibook.exe.
John Cork