Is necessary to import manually the stems before running TTFAutohint?

kupfers

Göran, is your “three minutes quickie-test” still online? Why not post a link, then we can all compare ourselves.

Göran Söderström

You can just go to Google fonts, see my last screen shots... same result.

Mark Simonson

Is it possible XP is applying some sort of "force bold" to the bold weight?

Mark Simonson

Could be a font meta-data issue:

http://typophile.com/node/95303

I'm not arguing against manual hinting. I'm just not convinced that it's a hinting problem in this particular instance.

PabloImpallari

Indra,
All the released and source files can be downloaded from here:
http://www.impallari.com/projects/overview/domine

And you can easily test any font on the drag-and-drop testing page, here:
http://www.impallari.com/testing/ (texts are editable).

Mark, Göran:
Thanks both for your comments and screenshots.
XP/Firefox and XP/Chrome seem to be OK, but XP/IE seems to have a problem.
I will investigate further. If it's a metadata issue we will fix it soon and update the files.
Thanks again.

Göran:
Which version of IE are you using? v6, v7, v8, v9? That info will help. Thanks.

Jason Campbell

"@jasonc
In this case I think it’s actually the TTFautohint tool that creates this problem. Without any hinting at all, I don’t think the dieresis would become a macron like this. "

Well in Cleartype, there's nothing that TTFautohint could do to change the rendering. Cleartype (for all intents and purposes) ignores any hints in that direction. It's only using the outlines.

I'm not saying that TTFautohinter is better than manual hinting, I don't think that Mark is either. FWIW, I've been manually hinting TTF fonts in VTT for over a dozen years.

Mark Simonson

.

Thomas Phinney

I've manually hinted a couple of big fonts in VTT. I'm also a ttfautohint backer (both personally and corporately at Extensis).

I have never heard anybody suggest that ttfa is better than manual TrueType hinting. That would be a silly statement. I do think that ttfa is among the best autohinting I've seen. It is also dramatically smaller than some alternatives, such as FontLab Studio autohinting, which is also important in some contexts.

I think ttfa is already very good, and with just a few more improvements ttfa could be clearly the best publicly available autohinter out there. That's a good thing. Nobody is going to hand-hint all the fonts already in existence, but if their rendering could be improved across the board, that would be a great thing.

If some folks think that better autohinting could allow them to avoid manual hinting, that's their call. I just think it will make more fonts more usable. There's already far more need for manual hinting than there is money, time or skilled people to do such hinting. That's not going to change any time soon.

T

Göran Söderström

I have never heard anybody suggest that ttfa is better than manual TrueType hinting. That would be a silly statement.

I have heard it a couple of times from the ”Libre-fonts-supporters”, but this is the only “proof” I can find:

(This is the only reason I even bother to discuss this at internet forums. I should stop though...)

Jason Campbell

FWIW, I think Claus's point in that thread was more the "bang for your buck" idea: that hinting is not seen at all on macs and less than half of it is seen in Windows/Cleartype. But he's also being cute, and perhaps... oh, I should stop there.

Dave Crossland

Thomas, what do you think those few improvements are?

EbenSorkin

I have to report that when I tried to re-create the link I thought I had seen between doing PS auto-hinting and results in TTFA I could not make it work again. I must have changed two variables and gotten a better result and misattributed the reason for the improvement.

EbenSorkin

On the question of "better" I do think the results can sometimes be better for some fonts simply because the results can retain more atmosphere or personality than is the case with manual hinting. It is not a robust enough tool to be able to hint very light or very bold fonts yet. Eventually it may be. There is a lot of potential in TTFA.

Paul van der Laan

Isn’t the point of manual hinting that you can determine *exactly* how much you wish to retain of the characteristics of the typeface?

John Hudson

The point of manual hinting is usually to enact explicit decisions about how you want particular shapes to look at particular sizes within the limits of particular rendering environments. So that might, in some situations, involve deciding how much of the characteristics of the individual typeface you wish to retain, but it is more likely to involve the opposite decision process: deciding which characteristics you want to sacrifice in order to achieve a more legible shape in particular circumstances.

Paul van der Laan

I agree with your comment, John. And it confirms my experience with this subject so far, that at its core, TrueType hinting is very much about making decisions that are design related. And therefore the best results are achieved when the hinter is either a type designer, or at least someone who can recognise certain design features of a typeface.

So it strikes me as odd that someone thinks that an autohinted font can retain more characteristics, given that the autohinter is completely inflexible, in comparison to a manually hinted font where every aspect of its appearance has been consciously made.

Göran Söderström

Thank so much Paul and John (!) for joining this discussion.

EbenSorkin

It does sounds odd. I can understand people being puzzled. And I agree that making deliberate decisions is absolutely the "meat" of design itself.

The reason I make this claim is that at the core of the original philosophy of hinting was the notion that it was desirable to make stems and other features into whole pixels. This made absolutely perfect sense in the era in which the technology was invented e.g black and white screens and very low resolution. As resolutions have risen MS's rendering engines have been getting more and more relaxed in their interpretations of shapes. This is reflected in the fact that in VTT you can target a rendering engine. This relaxation means a slow embrace of grey. This grey is where the personality of the type seems to live from what I have seen. From what I have seen manual hinting still keeps too much of this stiffness and admits too little grey. This seems very obvious to me even as I acknowledge that it is a kind of subjective observation.

If people want something other than Georgia and Verdana it will be because of personality and this is why I think TTFA is exciting. TTFA seems to let more of the personality through. In some ways this is a bit like saying Apple's Quartz engine lets more personality though which I also think is true.

The thing about TTFA is that it seems to embrace the grey more than manually hinted fonts.

As I have said there are some real limitations to TTFA. It doesn't do mixed scripts like chinese and latin yet. It doesn't do very bold or very light yet. There can be problems with i and j ( but there is a simple work around).

The main virtue of the process is that you can make something that works well not by hoping at the last moment that your design can be hinted into shape at a particular PPEM but you can instead design iteratively and test very very rapidly. This means designing into the tendency of TTFA. If you designed for VTT you might adjust the glyph shapes based on experience but the loop between gaining that insight via experience and applying it is much much longer.

Given huge resources I would be happy to use manual hinting but without them I am grateful that TTFA exists and that TTFA is an extremely design friendly tool.

I think that a hybrid in which you can get a TTFA result and then adjust the glyphs that are imperfect would be ideal. That might mean an 90% result with 10% to alter manually. I hope we see this emerge one day soon not least because it would mean more independent designers will hint their types due to the reduced scale of the job.

Anyway, I hope this clarifies my thinking/perception.

Jackson Showalter-Cavanaugh

That's why I set all my body text to #303030 instead of #000. More personality.

EbenSorkin

Silly.

Göran Söderström

Hahaha, this thread will go to history.

Paul van der Laan

I think it is not manual hinting that is too little acknowledging the grey, but rather the decision whether to hint or not to hint, that is already doing that. Even TTFA makes automatic decisions about stems and alignments.

If I would want to favour the “uninterpreted” shape of my typeface in a web browser, then I would just serve unhinted PostScript fonts. No ClearType jaggies to worry about, as Windows will always render them anti-aliased.

Jason Campbell

It's not manual hinting that is not allowing for grays, but the decisions of the hinters. In many cases these fonts were hinted for environments that benefited form a stark contrast. But it's the hinter who made that choice. There are strategies for manual hinting that allow for the hints to take advantage of the "gray" levels available. In fact most hinters I know of use these methods now, they aren't pushing things to look so "1998".

John Hudson

The history of rendering fonts on screen over the past twenty years has been a kind of dialectical struggle between legibility -- i.e. presenting the most easily recognisable character shape at given size and resolution -- and fidelity, i.e. preservation of the graphic characteristics of individual typeface designs (most of them originally intended for print media). Unfortunately, much of the propaganda of this war has tended to pretend that the struggle is about something else or, indeed, that technologies that increase fidelity are actually all about legibility. As soon as antialiasing was introduced, software makers made a commitment to improving fidelity, to making fonts on screen look like individual typefaces instead of generic character shapes expressed from pixel bricks. Thereafter, legibility was something that had to be added back, undoing the compromises of stroke density loss, fuzzing and alignment that unrestrained antialisasing produces.

Hinting is a weapon that has been used in different ways at different points of this struggle, and I don't think it is possible to generalise about what the purpose of hinting is. Indeed, one of the brilliant things about TT hinting is that it is a fully formed programming language, which means you can do things with it that someone else might not have thought to do. [Augsburger Initials!] In the main, though, I'd say that hinting has generally been employed on the side of legibility more often than on the side of fidelity, and this is why it remains important, albeit for some writing systems more than others, and for some rendering situations more than others.

The synthesis that eventually resolves this struggle is, obviously, higher resolution screens. But that's not a panacea, and my suspicion is that manual hinting won't completely go away for some time yet: it will simply become more and more specialised, focused on edge-cases rather than core environments.

EbenSorkin

I agree completely. In this sense it is clearly manual hinting that has the advantage in edge-cases if by that you mean the few glyphs that don't resolve well by auto-hinting or look funny at a specific PPEM. BUt was that what you meant?

TTFA is one very fast approach or flavor. It definitely isn't good for everything. And it doesn't let you solve problems except by changing your design. I hope this is fixed and that more edge-cases are resolvable by something other than re-design.

John Hudson

I was thinking of project-scale edge-cases: not so much a few troublesome glyphs amid others that do okay with autohinting, but things like making e.g. complex scripts with vertical stacking legible at 8pt on some particular device.

The Nirmala UI fonts for Win8 are an example of this kind of edge-case in that the technical constraints were tighter than for most fonts for screen use today because the lower target resolutions were still around 96 ppi and the primary sizes were 8pt and 9pt. We ended up flying Fiona Ross out here twice during the hinting so that she could advise on which stems to shift up or down at what sizes to produce the most legible forms for Indian readers. Meanwhile, we were also relying on hinting to squeeze extenders and marks into the limited vertical space permitted by the UI metrics.

That's the sort of thing I had in mind in terms of edge-cases.

James Puckett

We ended up flying Fiona Ross out here twice during the hinting so that she could advise on which stems to shift up or down at what sizes to produce the most legible forms for Indian readers.

Wow.

PabloImpallari

Hinting is a weapon that has been used in different ways at different points of this struggle, and I don't think it is possible to generalise about what the purpose of hinting is. Indeed, one of the brilliant things about TT hinting is that it is a fully formed programming language, which means you can do things with it that someone else might not have thought to do. [Augsburger Initials!]

Wow... those Augsburger Initials are awesome!!!!

EbenSorkin

RE: Paul comment "If I would want to favour the “uninterpreted” shape of my typeface in a web browser, then I would just serve unhinted PostScript fonts. No ClearType jaggies to worry about, as Windows will always render them anti-aliased."

I am not suggesting that is desirable to leave letter-shapes uninterpreted. What I am saying is that Windows has been letting its grip on interpretation relax a little with each new rendering engine. This means a bit more grey each time. With each relaxation there is more grey and less insistence on contrast. Along with having more pixels to use this loosening has let more personality through. I think this is desirable. Also to the extent that TTFA lets me soften things a little if I want, I appreciate that.

I don't know if the small foundries we will ever get it but if we were able to hint over the top of an auto-hinter as capable as TTFA it would be a great boon to us and to the readers of our fonts.

Deleted Account

Yes, I know.