Parameters to reproduce first figure about msdfgen HOT 8 CLOSED

Oh yes, you're right. Never paid attention to this detail.

That's called in ink trap!

from msdfgen.

Excellent: https://en.wikipedia.org/wiki/Ink_trap

from msdfgen.

Unfortunatelly, my method only improves corners and just like monochrome distance fields, doesn't handle well thin strokes that only span 1 or 2 pixels in the distance field. In the example, I used Arial Bold, which has thicker strokes. You should be able to reproduce it using the arguments in example.bat. But I admit the example is a bit biased, since 16x16 is almost never a sufficient size for a glyph's distance field, and I would strongly recommend choosing a larger size.

from msdfgen.

Thanks for the explanation, I wanted to be sure to have understood usage of your code.

There might room for improvement on the technique, maybe on the shader side ? If you look at this thread (you will realize we were waiting for your code) you'll see some possible tips: rougier/freetype-gl#110 and rougier/freetype-gl#97.

from msdfgen.

I'm afraid this cannot be fixed by any fragment shader (not to mention the test renders aren't generated with an actual shader, so it can't be a problem with fwidth or something like that), it's just the nature of distance fields. However, you can take advantage of the multiple channels to represent a thin stroke as the intersection of larger areas. You can do this by just using a different edge coloring strategy:

a_thin.txt (pass this as -shapedesc)

Although you would need more color channels to simultaneously also take care of corners, and I suspect certain configurations of multiple thin strokes would also cause problems.

from msdfgen.

I definitely need to read you paper (any tentative date ?) to get a better understanding of your technique.

By the way, I've been playing a bit more with the Arial font using different settings but in most cases, the upper pat of the internal left oblique line of the A letter seems to be curved (compared to the right) one. Could this be some kind of bug or a side-effect ?

msdfgen msdf -font Arial.ttf '0x41' -o msdf.png -size 32 32 -pxrange 4 -testrender render.png 1024 1024

from msdfgen.

Actually, I'm pretty sure that's exactly how it's defined in the Arial typeface. I have just tested that Firefox renders it the same way.

from msdfgen.

Oh yes, you're right. Never paid attention to this detail.

from msdfgen.