freetype
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Freetype] Comments on quality


From: David Turner
Subject: Re: [Freetype] Comments on quality
Date: Tue, 08 Jan 2002 15:51:03 +0100

Hello Artur,

Artur Zaprzala a écrit :
> 
> I installed freetype 2.0.6 (compiled without any modifications) on
> RedHat 7.1.
> 
> The first attachment is a small screenshot of:
> ftview -r 100 10 comic.ttf
> (comic.ttf is Comic Sans MS Regular v2.10)
> The letters P, b, c, d, i, k clearly (and ugly) stand out (are too thin).
>
> However, at 9pt A, M, N, V, v, w, x, y, z stand out (are too thick).
>
> Similar irregularities are noticable from 8 through 12 points (at 100dpi).
>
I know about the "Comic Sans" rendering problems in the auto-hinter, they
mainly come from the fact that the glyphs in this font are much more
irregular than a "classic" font, and the auto-hinter has some trouble
with these for now. One solution is to turn the bytecode interpreter..
 
> After recompiling with TrueType bytecode interpreter enabled, everything
> looks perfect (wow!), but there are exceptions :-(.
> The second attachment shows letter X (comic.ttf, 10pt, 100dpi, ftview),
> which partially dissappeared.
> When increasing and decreasing font size in ftview, I noticed that the
> position of letter Y relative to the baseline changes slightly from size
> to size.
> 
Uf you're using the TrueType bytecode interpreter, the problem could come
from buggy fonts. Disappearing diagonals is actually pretty common with
the MS Core fonts, because they've been hinted for monochrome rendering
only and there is nothing we can do without the ability to change the
font files themselves !!

Try disabling anti-aliased rendering for this font for "low" character
pixel sizes to get proper results. I didn't notice anything strange with the
position of "Y", I'll try to take a harder look though..

> And now Type1 fonts.
> When examining font Nimbus Sans L Regular v1.05 (URW) from
> urw-fonts-2.0-12.noarch.rpm with command:
> ftview -r 100 10 n019003l.pfb
> The letters AKMNVWXYZvwxy look to thick.
> 
> I could notice similar results with other fonts too.
> 
> What do you think? Is it buggy gcc, font, freetype or everything is ok?
>
It's certainly not a buggy gcc. It is my experience that some of the
Nimbus fonts available have pretty buggy hints (I mean, really stupid
hints in them), and I wonder how T1Lib deals with them.. For these fonts,
the auto-hinter performs much better, but it's difficult to tell Xft
how to choose the hinter..

I generally simply avoid them and use better fonts (e.g. the Lucidux
font family, the MS Core ones without "Comic Sans"). There are also
fonts that FreeType's Postscript hinter doesn't render too well too,
but this will be improved as time passes.

Improvements tend to be slow and difficult because any algorithmic
change routinely break the rendering of certain fonts while improving
others.. :o(


Note: I'm using 2.0.6 exclusively on my laptop and home machines,
      _without_ the TrueType bytecode interpreter and the result
      is very pleasant as long as you avoid the ugly fonts..

      For now, the things that still annoy me the most, and which
      will receive my attention in the coming weeks, are:

        - advance widths of certain auto-hinted italic glyphs
          (like "l", "i" or "n") is too short. This is
          unfortunately something a bit delicate to fix

        - accented characters don't display that well, depending
          on the font used. This is also a pretty hard topic

        - serifed fonts are still relatively ugly, though
          "Times New Roman" and "Georgia" begin to look good
          when used in Konqueror..


As always, improvements will be slow and incremental. But compared
to the first version of the auto-hinter, I'm 


Cheers,

- David
 
> --
> Artur Zaprza³a
> 
>   
> ------------------------------------------------------------------------------------------------------------------------------------------------------
>  [Image]  [Image]



reply via email to

[Prev in Thread] Current Thread [Next in Thread]