ttfautohint 0.8 screenshots with OpenSans

vernon adams's picture

I have made some comparative tests with OpenSans and ttfautohint 0.8. with WinGDI Cleartype and Greyscale.
Tested original font (OpenSans regular, Light, Extra Bold), plus - unhinted, ttfautohint with 'taller xheight', and regular ttfautohint versions.

The most significant issue remaining is the rendering at small text sizes with bolder weights, e.g. ttfautohinted versions can't yet emulate the finesse of a human manually hinting chars such as 'a', 'e', 'g' etc with such few pixels to play with. However, to my mind in the medium size text ranges i can see characters rendered a bit 'better' by ttfautohint ;)
Another gripe i still have is that ttfautohint tends to stretch the xheight by 1 pixel at particular small size instances. In my screenshots this happens noticeably at 12px, where the manually hinted x-height retains better proportion.

Suggestions for alternative fonts to test please. Manually hinted to 'industry standard' of course.


jasonc's picture

How about the Ubuntu Sans fonts hinted by Dalton Maag?

Jason C

hrant's picture

{To Follow}

Richard Fink's picture


I know, of course, what you're talking about when you describe the 1px "stretch".
That doesn't bother me all that much as long as it's consistent.

What's plagued me - and stopped me from using ttfautohint for release fonts except very, very occasionally - is the "shotgun" effect. At some sizes, there's pixels drop out that makes the letter look like a victim at the Saint Valentines Day Massacre. (google it, bro'.)
I've been neglectful and not taken screen shots. Do you know to what I'm referring? Have you seen it?

It tends not to happen with heavier weights.

Alas, we're at version .8 and still there are killer problems. I'm discouraged.


Té Rowan's picture

Oh, fiddle-faddle, @Richard! It isn't time to start doing the wowhawk until version 0.99!

blokland's picture

Reynir: ‘It isn't time to start doing the wowhawk until version 0.99!

I agree. In my opinion ttfautohint is an amazingly convenient tool for (batching) autohinting. My experience with developing software is that the closer one comes to the official release, the more difficult the problems that have to be solved become. And after seeing what has been improved in ttfautohint from the early versions on, I am pretty convinced that the resulting 1.0 edition will be impressive.

Concerning problems with initial software versions, I quite well recall the ATypI conference in Rome in 2002. We just finished one of the first official releases of DTL FontMaster and the focus during the development was on the Windows version. For Rome a Mac OS 9 version was compiled and I took it with me on my laptop, barely tested.
        Demonstrating it under Classic became actually a disaster, because the modules crashed constantly. If I remember correctly, there were some problems with a couple of MS libraries, and later I found out that a required temporary directory was missing at root level. Especially I remember that Yuri was quite relaxed after I gave him a personal, highly buggy FM-demonstration.


lemzwerg's picture

@Frank: Thanks for the kind words!
@Richard: Version 0.8 doesn't mean that there will be 0.9 and then 1.0. It will rather be 0.91, 0.92, etc. :-) I agree that it would have been better to start with version 0.01...

Additionally, I would be glad if you can send me a snapshot of the font which behaves as you've described (together with a short description of your rendering environment). It might be a systematic effect, but it also might be simply a bug.

For private reasons, I wasn't able to intensively work on ttfautohint recently. A new version will appear in the next few days, this time with only small improvements (mainly on the documentation side, and with reduced bytecode size).

Té Rowan's picture

Oh. Guess I thought you were zero-blanking the subversion number rather than counting version-and-decimal.

Syndicate content Syndicate content