ttfautohint - latest - autohinting Truetype Fonts

vernon adams's picture

I have upped a few screenshots from fonts auto-instructed with Werner Lemberg's very latest 'ttfautohint' tool.
It's still in development but it's covering ground very fast. ttfautohint instructs a Truetype font for web use in a matter of seconds.

for some screenshots comparing original hinted fonts with versions auto-instructed with ttfautohint, see -

mike_duggan's picture

hi John, thanks for the tests. The Greyscale with x-hinting looks pretty ropey to me, bad spacing, uneven stright stem v round weight distribution, heavy diagonals , spacing problems. I understand there is a trade off between this and more blurry stems, but you do gain more consistent spacing and typographic color with y only. Understanding that tradeoff, and then factoring in the work that is needed to fix the x+y version, might lead to a decision to only do y, which for some fonts may yield results that are acceptable.

Frode Bo Helland's picture

Manually, by an amateur:

Frode Bo Helland's picture

And a sans, a more relevant comparison perhaps. Sorry about the bg.

mike_duggan's picture

as John mentions the approach will depend on the weight and style and intended use of the font, but if there is enought weight, results can be *'ok' using just ydirection hinting for greyscale. see this quick example of Georgia, 11 point, top unhinted greyscale, bottom y-hinting only.

*We were never happy with the contrast we could achieve with greyscale and hinting at the lower test sizes, hence the GASP table. Hinting for black and white and editing the GASP, is one approach, for webfonts on XP systems.

If the outlines are lighter as in the sans examples above, then reducing the stems to one pixel, which is more or less the same as hinting for black and white, will necessitate hinting to fix spacing problems, and also to fix the balance in color between the straight stems and the rounds, and diagonals. Also as a side note, the Italic sans above, probably does not use x-hints right?

Frode Bo Helland's picture

"the Italic sans above, probably does not use x-hints right?"
No, it does, but none of the stems are locked to the grid.

dberlow's picture

FF: No, it does, [ probably not use x-hints ] but none of the stems are locked to the grid.

If x hints on italics, which have no grid to lock x stems to, then what do they lock to, or even do?

Frode Bo Helland's picture

They aren't locked to anything but themselves.

Yet another take on Win XP greyscale. Maybe the reason hinting tutorials are so hard to come by is the large number of possible ways to the goal.

pvanderlaan's picture

Maybe the reason hinting tutorials are so hard to come by is the large number of possible ways to the goal.


dberlow's picture

>They aren't locked to anything but themselves.

ohhhhhh. so... I hope you are instructing from the first point (which I also recommend you instruct no be rounded), to the second stem – you are doing a TT MDRP instruction with the booleans set thus:
m = not moving reference
r = Not rounding!
M = using the minimum distance value (default is 64/64ths of pixel), which keeps it 1 px at smaller end of range), and you can tinker with!:)
Bl = an instruction on a black feature (stem)

So, the only instruction of TT that's taking effect, (and I think this instruction is not going to work on Windows anyway), is the the minimum distance as qualified above.)

There is a lot more could be done, but unless you are hinting for aliased rendering at the smaller sizes in Freetype, or willing to trus MS who says if you spend a million hours you can perfect fonts with CT rendering), x hints on italics are seriously problematic to the point of "why?"

Frode Bo Helland's picture

Don’t ask me, David. I’m just testing some new approaches. Few others seem to. And, ouch – looking at these samples on a Mac reveal some serious gamma issues, in addition to the spacing issues I’m already struggling with. I still have about a thousand miles left on this path. If an experience hinter needs 80 hours per font I need some more experience.

Richard Fink's picture


"If an experienced hinter needs 80 hours per font I need some more experience."

I'm curious as to where that "80 hours" came from and how accurate it is. I read "40 hours" on, I think, Peter Bilak's blog. I'm just interested in how long the average manual VTT hinting job takes for, say, the WinAnsi set of characters.


dberlow's picture

>"If an experienced hinter needs 80 hours per font I need some more experience."

80 hours, for what kind of hinting? I'm not certain that's what I'd spend, even on an aliased font. Antialiased hinting takes less time per glyph and fewer glyphs need the kind of details to be hinted versus aliased hinting. 40 seems more like a reasonable estimate for hinting 79-80 antialiased glyphs without too much fuss beyond that count on adjacentcy issues. This also assumes some help from one or another of the auto hinters.

On the bright side, if you are spending 80 hours on any kind of hinting, you are getting more experience. And assuming that 60 of those 80 are getting you more experience than you need, then the next job will come to an early conclusion, perhaps.

Frode Bo Helland's picture

Uhm, yeah. Gaining experience. I haven’t timed my efforts yet, though. NB: Caps are a bit of a mess still but lowercase is pretty rad! (Review on a Win PC for correct gamma.)

Syndicate content Syndicate content