EB Garamond

georg's picture

Hi Typophiles!

I’m working on a digitization of the font shown on the “Berner Specimen”.

Of course I know, that this must be at least the n-th Garamond, but until today there is no such font available under an opensource licence, so I decided to try to fill this gap.

I want it to be as close to the original as possible but of course I make compromises where necessary to keep it usable. In the end (which might be far from now), I plan to have a family with different optical sizes and so on. For now it’s text size and roman only (and some first tries with the italic). Latin and basic cyrillic are present, greek will follow later.

I really would like your critique and comments about the font. You can get it from www.georgduffner.at/ebgaramond/EBGaramond.otf

Thanks, Georg


Since end of march, EB Garamond is available from Google webfonts for embedding.

Specimen.pdf88.24 KB
bojev's picture

Downloaded font from link - found strange jog or bump in top of lower case e.

georg's picture

That’s wierd, I’ve checked my local copy and the source and couldn’t find that bump. I’m working with FontForge. Can this be a problem?

By the way, I forgot, that the source files are available on github: https://github.com/georgd/EB-Garamond

JanekZ's picture

Isn't it a bit too light for text?
I don't like these flat acutes in UC. What is (probably) adequate in French looks odd in Polish (locl will resolve this problem).
Great job!

georg's picture

The font is quite light indeed, but it is stronger than the original. IIRC, AGaramond is about 5% stronger, the original would land at about 10 points below my font (vstem-width 60 instead of 70).

For the acutes, I have been thinking of polish kreska all the time with LC, but with UC forgot about it. I’ll try to repair this.

Pieter van Rosmalen's picture

The strange 'bumps' are in almost every lowercase glyph…


georg's picture

That’s very weird. I can’t find any bump in the letters. In the sfd file as well as in the otf, the top of the e is at 219|414 in one single point. What program are you using? Have you found bumps in other fonts produced with Fontforge? Might I have missed a setting?

georg's picture

If it helps, I have converted my sources to ufo format: http://www.georgduffner.at/ebgaramond/EBGaramond.ufo.zip

Té Rowan's picture

I do not see that dogtooth either - using ttfproof on a Linux box on freshly-downloaded fonts, both from Github and from georgduffner.at.

Pieter van Rosmalen's picture

I see the bumps when I open the font file in FontLab. The bump are there in all FontForge generated fonts I open, not only yours.


JanekZ's picture

FLS on Win: (I rounded the coordinates to Integer)

"One node selected" !!

This node has notInteger coordinates, maybe it is the reason?

Case resolved.
BTW Fontographer 5 makes the same error.

georg's picture

New font version should have no more notInteger coordinates and a new UC kreska together with polish locl lookup. Better like this?

Pieter van Rosmalen's picture

Yes, no more bumps.


Jens Kutilek's picture

It seems the difference you see with points with fractional coordinates is that in the UFO, the decimals are cut off, while when you open the OTF in FontLab Studio, the values are rounded to the nearest integer.

Original: 219.7|413.5
UFO: 219|413
FontLab: 220|414

mjkerpan's picture

Very nice start. At least on-screen (in the specimen and on your website, which I'm guessing is using EB Garamond through @font-face) in Linux, it looks pretty darn nice already. Beyond offering encouragement, the only recommendation I'd make at this point would be to consider also support SIL's Graphite smartfont technology in your font. While it only works with TTF fonts, it has the advantage of being the only Smartfont system supported by OpenOffice and LibreOffice...

cuttlefish's picture

With fractional point coordinates, a cumulative decimal place rounding error can displace the endpoint from the position of the beginning point, thus causing this jog. in FontForge, Element>Round>To Int snaps the point to grid for a selected point or all points in the glyph if none are selected, and solves this displacement on the endpoint.

georg's picture

Thanks for your comments so far, I had overseen the Round>To Int in the menu.

This is, what latin looks like at the moment:

This is the new kreska:

Sorry for the delay with posting images, I had some troubles with the GIMP…

@mjkerpan: Graphite would indeed be interesting at least as long as there is no proper support for OT in Open/LibreOffice, but I don’t have time at the moment to learn this language. If you or anybody else would prepare the gdl file, I would be happy to include this technique.

fujito's picture

Very nice to read. Keep up the good work. My suggestions:

- some parts are to dark for me: The serifs of the V or tail of the Q for example
- you use top and bottom extrema points, why don't you use also left/right extrema points?
("add extrema" in fontforge, and delete manually some redundant points (or "simplify")
- umlauts have same height as the i dot, I would lower them, also cdot has a too high dot
- generally spoken: diacritics are bad positioned

georg's picture

Thank you, Linus!

V, Y, X, Q are now lighter, dieresis and dotabove are lowered. Diacritic positioning is indeed not very easy for me and I’m not yet comfortable with the outline of circumflex and háček.

About the extrema: I had already employed the "add extrema" button after marking all glyphs with ctrl+a. Seems like this wasn’t enough. What’s more, I still don’t fully understand what the extrema are good for, as the curves are already defined by bézier control points (yes, I have not the least graphical education), Nevertheless, they should be in most glyphs by now.

georg's picture

I’ve now redone many diacritics. After playing around with the positioning, this is the current state:

What do you say?

litera's picture

At this size carons on /čšžř/ look quite tall for my taste... But not too big so I guess they could work at regular text sizes.

Just an observation. Shape of /s/ seems to fall a bit to the right. Just a tad bit.

georg's picture

Oh yeah, the s, my dearest letter.I tried to correct this:

For the carons at text size please have a look at the czech sample in the updated specimen in the first post.

litera's picture

I've checked your PDF. One think that I've seen is the different placing of caron on /ě/ and /ž/. On the latter it's positioned higher. Make sure it's the same on all letters otherwise it becomes very distracting for readers. But otherwise I still think they're quite tall on lower case letters. I don't know if you've made then thinner later, but in the PDF they also seem to be quite heavy.

Regarding the shape of /s/ it's still a bit to the right. Draw a line vertical line over it and you'll maybe see it better. Or have two of them side by side: one in original form and the other horizontally flipped. If the flipped one looks leaning to the left, it means the original is leaning to the right.

But if you don't change it it'll probably be just fine because in text it looks quite fine. I can see it others may not. So it's not that off any more. It's in acceptable range. The last change makes it even more appropriate.

P.S. I really like what you've done to the small caps Eszett/Scharfes S. Very slick with the top corner.

georg's picture

Now, the /s/ won’t fall over any more.

Also, I have reduced weight and hight of the háček, as you can see (I’ll do the same for circumflex later).

The háčeks over e and z do have the same height. I do diacritic positioning by anchors so the vertical position is transparently controllable with feature files. If you see a difference it’s either a presentation problem (screen, hinting or the like) or an optical illusion. If the latter is the case and is reproduceable for others too, I should think about changing anchor positions for some letters.

As for the smcp Eszett, this glyph was the beginning of my learning to draw letters. As it is missing in most of the typefaces I started to draw it for different ones. It’s not a very useful one (seldomly used, and if you use it, people will stumble over it), but I really love it :) The one for EB Garamond was drawn very early in the whole proces and barely touched since then (about 9 months ago) so one might deduce from it some learning proces :)

BTW, the specimen has slovenian text now as well ;)

litera's picture

It looks better already. :)
I hope someone else will reply as well so it won't be just my opinion on this. Do check other fonts as well. But I guess this looks better.

Slovenian only uses čšž so I guess we're covered yes. :)

georg's picture

Oh, how stupid. Of course it is an optical thing. The diacritics need some kind of vertical kerning. I’ll reposition the anchors.

mustail's picture

Your font looks amazing, congratulations. I use URW-Garamond for my LaTeX documents, but I find the dots to be too high above the letters for my taste. I am happy also that you have a wide range of glyphs. I wonder if you were considering including letters with a dot below, such as Ḥ Ṭ Ḍ Ṣ Ẓ . These glyphs are necessary when transliterating Arabic and Persian into Latin, but not all Garamond fonts (such as the one provided with MS Office) have these.

georg's picture

Thanks for your congratulations!

Shortcomings of URW Garamond were in fact one of the reasons to start this project. Although I would like it, I’m not able to make fonts for classical (La)TeX, so you’ll have to use Xe(La)TeX or Lua(La)TeX with EB Garamond.

Enabling the Garamond for scholar use is very important for me, so the letters with dot below are in the font and since today they also show the dot:

flooce's picture

sry, 2xpost

flooce's picture

Did you consider to employ any method to ensure that the /f/ does not come to close to letters like ä,ö,ü,å ?
Some designers solve this with contextual alternatives where the f becomes slimmer if an Umlaut follows (e.g. Calluna), others try to solve it with ligatures.

georg's picture

Yes, this feature is included and should be active by default (calt). It replaces f/ff/ſ/ſſ by a shorter version before letters with ascender or diacritic when it doesn’t get ligated to it, and before space, symbols etc. This was actually one of the first opentype lookups i included in the font :)

Gary Lonergan's picture

An amzing amount of workcongrats I have eons to go before I even consider contextual alternatives

georg's picture

Thank you, Gary!
I think, your Dutch face would do very well with an f with wider arc as a contextual alternative.

Gary Lonergan's picture

Good idea you see I'm not thinking that way yet. I work on magazines and newspapers and wanted to avoid a kerning f because of the problems I've had in the past but contextual alternatives are the way to go.

flooce's picture

Thank you for the reply and the work you put into it. I didn’t notice my contextual alternatives settings were switched off. It is good to see this feature is taken up in such a promising font.

I had a look at some of Slimbachs fonts on myfonts.com, like Minion, both Garamonds, Adobe Caslon (Slimbach too?) and Warnock, and out of those only Warnock avoids the collision. I hope this feature will be made us of in future on a broader scale.

In any way, you can be proud of this interpretation. Thank you for your work.

georg's picture

Florian, danke für das große Lob, das ehrt mich sehr!

flooce's picture

I meant “will be made use of” like: will be implemented often...

Ach, ich danke zurück. But you know... who am I to tell.

Syndicate content Syndicate content