Typophile

Painting and/or Superhinting and/or ClearType - Relative Effort

Hrant H Papazian

Spawned from here:
http://typophile.com/node/18685

So along the journey on the eternal quest of making a font
look good (but to me at least never “the same”) both in print
and on screen, I have a question:

What’s the relative effort involved in getting this
outline “g” (slightly modified from Patria) to render
out this bitmap “g”, preferably robustly on as many
platforms and applications as possible.

One way is to make two separate fonts, one outline, one pixelfont*.
Another is to superhint, like the previous generation of MS core fonts,
but with the difference that grayscaling is enabled. A third is to -come
close, and in some ways exceed- using CT, using some level of hinting
(which I’m not clear on). For the purposes of this discussion I will add
that -if it makes a big difference- you can assume the top of the lower
bowl of that outline “g” to be flat -and even uniform in weight- instead.

* Or a raw bitmap font, in special situations.

So I’m wondering what are the relative efforts, as well as
the relative technical impediments in terms of delivery,
of the various methods. Thank you.

hhp

‹ all caps monospaced electro-mechanical printer type Five Volume Set ›
Hrant H Papazian

BTW, another relevant comparison here is to grasp the relative effort required in grayscale superhinting versus 1-bit superhinting (something that many people have regularly done). Basically to answer the question: Is it worth it?

hhp

David Berlow

“So I’m wondering what are the relative efforts, as well as
the relative technical impediments in terms of delivery,
of the various methods.”

watz the spec.?

Hrant H Papazian

Spec? Wazzat?

Maybe this: a TT font populated with “normal” outlines (leftmost “g”), but that renders out specific grayscale bitmaps (righthand “g”s) at certain PPEMs. My sample bitmap “g” is from an existing (but still in development) 14* PPEM serif pixelfont, which however has quite “abnormal” outlines (because without hinting that’s the only way to get specific grayscale bitmaps).

* Or 15 if you want to include a blank topmost row (although that’s a
gracious sacrifice of apparent size that’s been unfashionable for years).

So how painful would it be to superhint a normal outline font to render out bitmaps like that? One reason I ask is because I wonder why it’s never been done (AFAIK). Although maybe the Next Big Thing became ClearType so superhinting* was exiled from the Republic. And speaking of CT, would that be any better in terms of reducing effort? I’m assuming CT can render out “normal” (full-pixel), non-colorized bitmaps.

* Which interestingly however is not mutually exclusive with CT, so...

As for the cross-platform, cross-app wish, that might actually be pretty hopeless, seeing as how OSX mostly ignores hints. On the other hand, CT-style fonts might actually work when OSX’s subpixel rendering is enabled? What’s funny BTW is that the existing format of my pixelfont actually ensures consistent rendering across platforms! The only other way to do that these days seems to be to settle for Fabulous Full Fuzz on both platforms! :-/

hhp

hankzane

I don’t understand - why don’t you fire up FontLab or VTT and find out for yourself?

Grayscale hinting was done on Palatino, to some extent. I think Mike told me that, but ... I’m not sure. Perhaps I have diagonal hinting in mind. I believe Bistream Vera family was also hinted for grayscale rendering (by FreeType).

Hrant H Papazian

> why don’t you fire up FontLab or VTT and find out for yourself?

Because I’d have to spend a long time trying various things out
to find out what certain others could tell me in a few minutes.
But if I have to, I will.

When somebody puts up a font for critique, I don’t
tell them to go spend some years figuring it all out.

> Grayscale hinting was done on Palatino

Oh, wait, I think Simon told me that too, not long ago.
I’m so hare-brained these days I can’t even remember
answers to my own questions. I was supposed to check
that out...

> I believe Bistream Vera family was also hinted for grayscale

Can anybody confirm/deny this?

—

BTW, I wrote:
> I’m assuming CT can render out “normal”
> (full-pixel), non-colorized bitmaps.

But I’m realizing that in order to get that a font would
actually have to be... superhinted, and quite extensively!

hhp

Captured from web.archive.org on 24 Dec 2007. view this capture