@mike_duggan sounds like we have to set a tutorial on our websites for the visitors, so they can enable ClearType. the thing is we all here now that, but the visitor does not…
on Windows Vista and Windows 7, ClearType is the default rendering on the system. It may be that frode frank is running Windows XP, where the default is set to Font Smoothing, with the option to turn on ClearType.
So, in a lot of cases the customer does not need to choose ClearType
more info here
@siiFailing on font rendering quality is the bigger risk.
Nonsense. Everything favors the early entrant with the lowest barriers. Ugly, beautiful, whatever. Not everybody shares your allergic aversions to what constitutes "ugly" type. The uglier it is, the better I like it. I've told you that.
What counts more is that it's not the same old web safe mix. It's a different look.
Me:The baseline is: how do the intalled workhorse fonts look in comparison? You can't evaluate in a vacuum.
If the user is quite used to Arial and Georgia and Verdana looking like crap, they won't care less about some other font chosen by the web designer that likewise looks like crap.
frode, baby, bubby, what do those fonts look like on the machine you did your screen grabs on? "None", "grayscale", "font smoothing", "ClearType" - these are causes which mean nothing without the context of their effect on every font the user is seeing in the browser.
Rich: ‘Arial and Georgia and Verdana’ were all heavily hinted for aliased text rendering (i.e. no font smoothing at all, just square, monochrome pixels). In addition, Georgia and Verdana were expressly designed for screen reading in aliased and greyscale rendering (grid fitted, hinted, and spaced for screen). ClearType rendering ignores the whole pixel x-direction delta hints in such fonts, but retains all other aspects that make these fonts well-suited to text on screen. So whatever the current rendering environment, whether with no antialiasing, greyscale antialiasing or sub-pixel antialiasing, these fonts are still going to perform better than the vast majority of other fonts.
Further, as I demonstrated earlier in response to your ‘Ten Great Free Fonts Cross-Browser’ TT conversions, comparisons of potential web text faces need to be made at text sizes, not looking at half-inch high glyphs. If one is comparing display faces, okay, show them as large as you like, but what fundamentally matters is not whether big type looks pretty or ugly but whether small type is readable.
I'll meet up with my brother later in the week again. Perhaps I can get some screenshots of Verdana/Georgia then.
@mike_duggan You are right, it's enabled by default on win7 and vista, my concern is the nearly 59% with WinXP
@John Hudson I'm with you, web typefaces must be readable at any size, and that require manual hinting
Rich, I wrote my comment only because of your sarcastic response to Jens, that you have no idea what hinting is all about – yet in your explanation confuse hinting and subpixel rendering, which are two very different methods. But your blog post is almost three months old and I am sure you know much more now.
So whatever the current rendering environment, whether with no antialiasing, greyscale antialiasing or sub-pixel antialiasing, these fonts are still going to perform better than the vast majority of other fonts.
OK. So, Frode, you still there? Take a grab of Georgia and Verdana in that same environment and let's see how much better, if at all measurable, those fonts look.
Seeing is believing.
(Frode, frankly, I don't really have the time right now to boot my XP SP2 machine with IE6 and start playing with settings, let alone displays. Help me out, willya?)
Congruence is the issue, not some system settings.
Do you think computer users with older systems running older browsers are unaware that what they are getting is a degraded version of what they could get with a newer system running the latest version of an A-list browser? I don't.
If one is comparing display faces, okay, show them as large as you like, but what fundamentally matters is not whether big type looks pretty or ugly but whether small type is readable.
Not if it never gets small. I really don't think anybody is going to use a font like Tagesschrift for paragraph text. Fonts have "sweet spots". Certain sizes, measures, just work better than others. (Telling you that is absurd, but a lot of people don't know.)
Play the song "Begin The Beguine" in the key of C and then play it in the key of A Flat and there will be a very discernible difference you can feel, even though all the internal relationships - that is, the harmonic intervals, melody, et al - remain exactly the same. (Sorry to non-musicians for the analogy, but I assure you it's on the money.)
Computer screens, as a medium, would appear to make a strange demand on typefaces as opposed to print - the demand to look equally "good" at any size. I think this is unattainable. In fact, I don't even think it's worth the try. What to do about it, I don't know.
So anyway, to cap off, it seems to me you're saying that what type designers looking to cater to web designers should strive for is as close as they can come to the construction of the web safe fonts.
Sounds good to me.
After all the sturm and drang of last year over EOT/CWT and its main advantage being backwards compatibility with IE 6, 7, and 8, I would assume type designers would want to take advantage of that.
>>Failing on font rendering quality is the bigger risk.
I think you missed the point. I said rendering is a bigger risk than business model, as if you have good quality fonts the business model can be changed instantly, you can't fix a rendering issue instantly.
Should add that when I made that comment back on Feb 25, this thread was about FSI’s licensing model, and that remained the focus for another few weeks. I didn’t change the subject to rendering, the people looking at the fonts changed the subject. At the end of the day it always comes down to quality.
Rich: Seeing is believing.
What, you don't remember what aliased Verdana and Georgia looked like on your CRT monitor for all those years?
Here are some benchmark (representative) images of available renderings of these fonts in Vista: GDI aliased, GDI ClearType antialiased, and DWrite ClearType antialiased (the latter with y-direction greyscaling, subpixel positioning, and subpixel kerning). I can't make greyscale antialiasing showings at this size without hacking the font gasp tables: these are set so that greyscale antialiasing kicks in at larger sizes.
It's worth noting that screenshots of CT rendering may appear different on your computer than the local CT rendering due to gamma, resolution, and other factors. However, with that caveat, I suggest that any and all showings of web fonts intended for body text should be comparable to these benchmarks; most importantly, they should be at the same or similar size.
To me the ClearType samples look awful, because I see all colors of the rainbow in what is supposed to be black text on white background. (Maybe my eyes are weird with colors. But from what I have been gathering, many other people's eyes are, too.)
Aren't all “font rendering problems” in discussion these days actually created by the wysiwyg-philosophy? If we treat the screen as a screen – and not as an emulation of paper – all problems seem to disappear.
And as more and more of our content is produced and consumed online and on screens – what's the point of wysiwyg anyway?
All exits are closed, but we can always escape through the main entrance... :-)
as John mentioned, It's worth noting that screenshots of CT rendering may appear different on your computer than the local CT rendering due to gamma, resolution, and other factors. If you are using Windows, and ClearType that is?
I tried to make the same screenshot of DWrite rendering as John, but I just can’t. Subpixel rendering looks very weak (my example is second):
I used IE 9, but the same rendering is in FF also using DWrite.
First I thought that I have some strange settings, so I updated nVidia drivers, made sure there are no strange gamma settings, color profile was OK, ClearType settings had no impact on DWrite.
On the other hand, I could find a couple of images showing new ClearType, and there were also this weak colors of subpixel rendering – weak compared to previous version of ClearType.
So, has new ClearType really weaker colors of subpixel rendering?
I am looking at the screenshots on Mac OSX. But I see colors every time I look at ClearType on Windows, too. I find it more disturbing than Quartz, even though I dislike Quartz subpixel rendering for the same reason (colors). My favorite setting is simple greyscale anti-alias, if possible gridfitted. Maybe I'm just hypersensitive to colors?
The point I am trying to make is that many of the recent ‘problems’ with font rendering on the web come from using fonts and conventions from print (wysiwyg). From my (very peculiar) point-of-view, the issue of sharp screen type has been solved several years ago with bitmap fonts and pixel fonts for Flash. Such fonts don't look good on print, but with html/css one can easily specify a proper stylesheet for printed media, with proper print fonts.
>So, has new ClearType really weaker colors of subpixel rendering?
GDI and DWrite CT can be tuned separately (hence the multiple screens on the CT tuner), and by default DWrite uses a different setting to the default GDI setting (a result of much testing AFAIK) - try re-running the Windows 7 ClearType tuner.
Thanks Sii, it worked – the new setting just wasn’t instantaneously visible like in old CT.
First, sorry for responding to a second hand account of what you said. Shame on me for that. Nullius In Verba.
>At the end of the day it always comes down to quality.
Yes and no. Depends on the definition of quality. (Having spent years in manufacturing, I'm inclined towards the philosophy - and science - of Edward Deming.)More often than not what those within a craft consider the important "qualities" are not the most important to the bulk of customers.
Free is still free, after all. And price is a quality.
>What, you don't remember what aliased Verdana and Georgia looked like on your CRT monitor for all those years?
Is that what that strange, recurring dream I have is really about? Now I can seek treatment. Thanks! (Thought it was due to some of the pharmacological delights in which I indulged as a young bacchant.)
(BTW - Yet another reason why web fonts failed to gain any traction until recently was the predominance of CRTs. When the benefit is going from not too awful to different but still pretty awful, there isn't much of an impetus - even if the bandwidth had permitted it.)
Great screen shots, BTW. Hope you don't mind if I appropriate them.
Re: rendering with bitmap screen shots - as was pointed out to me by my FB friend Bill Hill some time ago, compressed raster formats like GIF, PNG, JPEG, fail to accurately re-create. You need BMP or TIFF to grab the full color spectrum. Seems only logical, and to my eye, that certainly seems the case. (True?)
But of course the size of the image is a deal-breaker when posting to a forum.
When you say "weaker", I'm sure you're right in some sense because the one thing I'm not seeing in CT 2 is the "fattening" of the font that I see on the Mac.
And I'm quite happy about that. Looks like better fidelity to me. And a much more readable font at small sizes.
My sincerest condolences! I think I would be less polite, if I had licensed FontFonts. The upgrade should not cost more than 20%. Everything else is usury.
>The point I am trying to make is that many of the recent ‘problems’ with font rendering on the web come from using fonts and conventions from print (wysiwyg).
Don't bother. ;-)
So, has new ClearType really weaker colors of subpixel rendering?
I realise, looking at Miha's comparison, that my examples may not be representative of default tunings in the release version of DWrite, as I made them with a rendering test tool. [So, Rich, perhaps best if you don't appropriate them, but try to get some renderings from e.g. IE9 or the latest FF pre-release builds.]
Verdana's alright, Georgia's alright, they just seem a little weird,
duh-render, duh-render, but don't give yourself away.
Are there one-size, one-platform benchmarks in web type? hmmmmmmmmmmmmmmmmm....
Or is it duh-render, duh-render, but don't give yourself a way?
Hard to decide with all the cheap tricks flying. ;)
Are there one-size, one-platform benchmarks in web type?
No, and I didn't suggest that there were. But there are one-size, one-platform benchmarks for comparing different text fonts at the same size on the same platform, which so far has not happened with all the rush to embrace the desktop publishing c. 1985 approach to web fonts.
>But there are one-size, one-platform benchmarks for comparing different text fonts at the same size on the same platform,
I think this is a thin slice of bench and not much of a mark.
For example, what if the comparison font is designed based on the ppm grid at which you want to compare a perfectly good text font designed based on another ppm grid? The benchmark font could look GREAT on its grid but CraPola on the grid of the font you want to compare it to... Is your font better than the benchmark, or worse? Or.... should all fonts be drawn to the grid of the benchmark? Are fonts with built-in leading of any amount, forever inferior at all text sizes?
And once you've seen these fonts on one platform at one size, and determined that they are equally brilliant or mediocre, What do you know about them as web text fonts?
>... which so far has not happened...
I think a number of people have lent little voices all along, and others have been making more noise based on tests such as this but older. The results then were that we need all the great digital font technologies we've ever had ('cept for maybe 1), not a lowest common denominator of two, to really solve the issues of web typography.
I needed most of these technologies to design a puny-human web-font-format bench, before I could begin to mark it ;)
Same computer, same settings.
Forgive me, I did a screenshot with cleartype enabled as well. Must have picked the wrong one.
Yes, Frank, there can be large differences in horizontal metrics between aliased and anti-aliased type of the same text on the same computer.
Mr. Hudson's 'benchmarks' e.g. do not show this because they are chosen fortuitously, or something.
Or is it something about the aliased type on the right? or the missing grey squares,
I take it you are referring to the darker grey squares on the grey background, David. They're merely a design feature (see www.mosaicsolar.com).
I'm really not a part of the discussion in this thread, which unfortunately seems a bit heated. I just wanted to know if there was a difference in hinting between the Web Fontfonts offered on Typekit and those they sell from Fontshop.com. Turns out it's only a case of my brothers computer having Cleartype (and "Greyscale") disabled.
All the best, and happy holidays!
Mr Berlow: Mr. Hudson's 'benchmarks' e.g. do not show this because they are chosen fortuitously, or something.<.em>
Or because they're Georgia and Verdana, and are designed to be close to grid-fitted stems and spacing -- at or somewhere close to the ppem size I show --, so don't vary in width horizontally between aliased and antialised as much as a lot of other fonts will.
Mr Berlow again: And once you've seen these fonts on one platform at one size, and determined that they are equally brilliant or mediocre, What do you know about them as web text fonts?
So far, you'll know that most of them are crap, and that the web is going to become a very ugly and difficult thing to read in the next couple of years.
Re. fonts designed for specific ppem sizes: a) obviously they should be compared at the size at which they look best and b) they don't worry me, especially not if you are making them.
Can I go back to calling you David now?