Meir, could you show a full alphabet setting? Since I don't read Hebrew, my only -limited- way of seeing what a Hebrew font is "doing" is to compare its structures to those of other fonts that I've been analyzing lately.

hhp

Alright, I've done some alterations that I believe improve readability a great deal, like "squaring out" the top left corners of the roundest letters, and also making the commas and parentheses larger, so readers would be able to differenciate them from similar symbols better.

OK, here's my (highly "theoretical"!) feedback:
- I'd increase the word space.
- Some of the sidebearings (like the Dalet's left) should be zero.
- I like your vertical proportions, I just wonder if it wouldn't be better to give the Lamed a 2-pixel ascender.
- Widths: I think it's better too wide than too narrow, I just wonder if width variance can't be used to diverge the confusable letters even more, like in the Bet versus Kaf.
- Alef: what about making the top-right arm shorter (coming out from higher up in the diagonal)? That would reduce the distracting symmetry.
- I really like the bravery of the Gimel.
- The wobble in the Zayin is valuable, but it's too jittery in small bitmaps - I'd try not bringing the down stroke back to the right after the left shift. So basically a flipped Vav (shorter by a pixel) with a flat top added.
- I wonder if the Yod needs to be stronger.
- What about giving the (non-final) Mem an ascending pixel?
- What can you do to make the Nun less like the Kaf?
- What about giving the Samekh an extra leftward pixel on top?
- Smart divergence between the Qof and the He.
- The final-Mem is too square. Hey, it is a square! What about curling in the left stem at the very top?

And a question: If you make a bold, you'll put the weight in the horizontals. right?

hhp

What wonderful feedback! Alright...

- I'd increase the word space.
Yes, it was very needed.

- Some of the sidebearings (like the Dalet's left) should be zero.
I don't agree. The terribly small size of the letters makes spacing crucial. Zeroing a sidebearing on any of the letters (I've tried numerous tracking combinations) yields very odd implications. Kerning would be super if I could get it to work... Haven't mastered that field technically, I'm afraid.

- I like your vertical proportions, I just wonder if it wouldn't be better to give the Lamed a 2-pixel ascender.
You're right on. I hate making Lamed "ascenders" (in Hebrew we call them "poles") too short, it just isn't natural. Look at this caricature from an old Hebrew calligraphy/typography teaching book:


- Widths: I think it's better too wide than too narrow, I just wonder if width variance can't be used to diverge the confusable letters even more, like in the Bet versus Kaf.
True, but that is not my main concern about these wider letters. I'm more troubled by the notion it makes the eye linger on them, trying to decode these odd shapes' meanings. I'd hate having to part with this incredible addition to my font.

- Alef: what about making the top-right arm shorter (coming out from higher up in the diagonal)? That would reduce the distracting symmetry.
I've made it assymetrical now, correspondant with the Tzadi, which I believe is a letter meant to be similar to the Alef.

- I really like the bravery of the Gimel.
Thank you! It stayed that way from the 2000 version.

- The wobble in the Zayin is valuable, but it's too jittery in small bitmaps - I'd try not bringing the down stroke back to the right after the left shift. So basically a flipped Vav (shorter by a pixel) with a flat top added.
Shifted the upper pixel back. I don't like the idea of Zayin being too narrow, I want more-or-less similar widths for the letters, because I believe it improves readability.

- I wonder if the Yod needs to be stronger.
I've widened the Yod, Vav and Nun-Sofit.

- What about giving the (non-final) Mem an ascending pixel?
That wouldn't be very wise, an ascending pixel would make it seem bigger than the other letters. I've taken off a pixel from the bottom, to emphasize its Mem-ity :>

- What can you do to make the Nun less like the Kaf?
I've added a pixel on the bottom right edge.

- What about giving the Samekh an extra leftward pixel on top?
Nah, I don't want the letters to be too square.

- Smart divergence between the Qof and the He.
Thanks! This has also been in the 2000 version.

- The final-Mem is too square. Hey, it is a square! What about curling in the left stem at the very top?
Oops! Got me there. Here, it's fixed now.

Again, thank you so much for your incredible feeback!

> Kerning would be super if I could get it to work...

Well, maybe you're right about the zero sidebearings considering kerning is not always available, but why are you having trouble with kerning?

> I'm more troubled by the notion it makes the
> eye linger on them, trying to decode these
> odd shapes' meanings.

Actually, you're right: at this size legibility is more important than readability. (See below.)

BTW, are they really "odd"?
Are they odd because of a conscious attempt by you or because of the lo-res?

> I want more-or-less similar widths for the letters, because I believe it improves readability.

Actually, width divergence is a big factor is readability (as long as no glyph is too narrow). BUT, legibility (different than readability) does benefit from "raw" size (= width), so I think a wide Zayin is smart, especially in the context of the Vav. BTW, I think the old Vev (and the old final-Nun) were better (more divergent from similars).

You know what? The confusing network in Hebrew is very strong and very pervasive.
I need to make a diagram, like the one I've done for Latin:
http://www.themicrofoundry.com/other/m&s.gif

What would be really nice though is some linguistic data - like at least letter frequencies, or in this case ideally a (long) list of the most frequent Hebrew words. That would help seeing which letters are most confusable in actual text, not just in a structural vacuum. I think John has some info like this, no?

hhp

> why are you having trouble with kerning?
I just can't get it to work... Dunno why.

> I think the old Vev (and the old final-Nun) were better (more divergent from similars).
Hmm. I'd have to consider that.

> a (long) list of the most frequent Hebrew words.
I might be able to help you out with that!
I have this ongoing on-line project on my Hebrew site called Patpetet:
http://www.oketz.com/patpetet/
It's the Hebrew version of "Blather", if you're familiar with it. Basically, users have been logging in for the past four years and feeding the database with all kinds of words. Most of these words are pretty much frequent (hey, they're HUMAN-generated! that's gotta count) and there's like 2000.

Anyways, I extracted all the words from the database onto this list:
http://www.sadan.com/hebrew_word_list.txt
(I've UTF-8ed it.)

Will that be okay? That diagram you have there is pretty useful. I bet it can be done for Hebrew.

Meir, this is a really nice piece of work.

Im not so sure i understand why you made the letters with the dagesh wider then the ones without it. Is there a rational to it?
in a thinner pei with dagesh you can make the pei not enter the the space inside. that way the dagesh wont connect to the pei's body.

I really like the alef. its extremly nice. the whole font has a good feeling to it, like the old concept font.

Maybe the old zayin was better. consider its aesthetic value.

I feel we stumble into the same problems in hebrew screen fonts. for example, the mem looks very much like the het. if you drop that pixel at the bottom of the mem you'll be left with a het. the subtle differance is also present in taf and het, tet and samech, and by some extension of the phenomena its also present in- bet and kaf & resh and dalet.
do you find this one-pixel diffrence as disturbing as i do? *sigh* i think its a major problem.
i tryed solving it by making all letters wider and more 'detailed'. i sort of droped the idea of having a mem without some sort of oketz, or a nice circle-shaped samech, etc'. well, you saw Kyoto. its not so much a solution to this legibility problem as an escape from it.

(argh.. meir, your english is good

Itay, the rationale behind the wide letters is that the overall shape of the letter is preserved, while positioning the dagesh mark in the exact middle of the letter's inner space.
Understand I'm having moral difficulties altering the Pei. :> It simply isn't what I'm trying to acheive in this font.

Answering Hrant's unanswered "oddity" question, the shapes of the wide letters are indeed odd, but only because modern Hebrew readers aren't used to wide letters, that are a norm in prayer books and such.

Being a Sevenet compliant, I'm trying to make it more or less wide enough and simple enough to be associated with the simple latin letters of that face.

The one-pixel difference is biasing, but in this occurance is also manageable. The base matrix for this font is 5x5 pixels, meaning 25 pixels give an overhaul 100% for a single, average letter. Altering one pixel would affect 4% of the entire composition - that is a lot! And when we're talking about its perimeter, that's almost 6%, and when we're talking about a single edge, well, now we're at 20%!



Oh, and also referring Hrant's question about making it bolder - I wouldn't dare making this specific font bolder, but when it comes to making heavy-weight Hebrew fonts, I either strengthen its horizontal strokes, or make both horizontal and vertical strokes bolder. I avoid strengthening only the vertical lines because it gives Hebrew letters a very fragile look.

(I hope it wont sound as if I'm attacking you. i want to debate about this because it bothers me. Because it bothers me, I might sound a little too aggressive. :/)

In a way youre right. Inside the letter the change isn

Okay, I totally accept what you're saying, Itay, but I'm reserving my right to agree/not agree with different parts of what you said. :>

My calculations does reflect a practical side of the letters, but optical illusions, much like you said, don't go by mathematical formulae. The change of one pixel greatly affects a letter's appearance, but the position of that pixel within the context of the letter (whether it belongs in a horizontal row or vertical column, for instance) is crucial.

I've made some changes to the font, but I'm not very pleased with them, so I'm withholding it from showing it to you guys until I'm run out of ideas. Remember, I'm keeping this one simple (hell, it's 5 by 5)!

(OK, back in the loop on this one.)

> http://www.oketz.com/patpetet/

Although I can't read any of it, I understand what it is. It's certainly useful, because it addresses words that are neither too frequent nor too rare. Is that an accurate assesment? If so, your list would be very helpful for kerning - thanks!

As for plain letter frequencies, do you know where I can get them?

> [Kyoto is] not so much a solution to this legibility problem as an escape from it.

That's a nice way of putting it! Some of our efforts in the field of legibility are just a release.

Itay, where can I see Kyoto?

--

As for the "one-pixel difference" issue, I have to mostly agree with Itay. You need the difference to appear structural, and sometimes that's possible with one pixel, but not usually.

hhp

> Is that an accurate assesment?
I believe it is!

> As for plain letter frequencies, do you know where I can get them?
That's trickier. Maybe that can be extracted from that list as well, if it's properly filtered and then processed. I'll look into it when I'll have time.

> the rationale behind the wide letters is that
> the overall shape of the letter is preserved,
> while positioning the dagesh mark in the exact
> middle of the letter's inner space.

I just realized something: why not make the dots two pixels wide in those cases?

hhp

Hmm... I guess it's possible. Although the outcome looks a bit weird to me, here's the latest version:



Here's an alternative, cursive version I've been working on:

I think the 1&2-pixel dots are actually working.

The cursive: it seems very interesting, but here I'm totally out of my depth (more like totally into my shallowness). I'd be interested ot hear what Itay thinks of it.

hhp

damn, its really nice. i especially dig the way the dagesh works in this alt version. the tzadi and pei sofit are very original.
some sharp corners look out of place to me. like the mem and the inner part of the shin. i think the whole font might profit if you totaly went for a soft look, and smoothed those letters.
the shekel sign could be softer too.
have you thought about making a two-part alef? its a probably less legible option, but it might look nicer in this font.
im not sure about the zayin. you might get it confused with the nun.
also, i wonder how the resh & dalet would work minus one pixel in width.

anywho, its an interesting font.. experimental.

hrant, ill post my Kyoto soon

This is the latest version of Concept Alternative:

Only minor changes have been applied, most glyphs remain the same.

So is this cursive style "orthodox" (pardon the pun)? I'm really wondering what natives (or even any Hebrew readers) think of it in terms of general acceptability.

hhp

Hebrew "cursive" style is also the Hebrew "script", it's simply how Hebrew literates write - you learn it at school when you're 6 years old. So I guess it's pretty "orthodox", yeah. :>

It isn't really used for book text or anything, though. It might be appropriate for side notes, wedding invitations or poetry (just like Latin cursive, actually).

OK, that's good news. So what would happen if a book used this style?

hhp

I'm trying to think of a book that used cursive letters, and all I can recall is a Hebrew translation of a volume of Japanese love letters. The text was handwritten by a calligrapher, so I don't know if that counts.


In any case, I believe that in order for "cursive" styled letters to be used in the context of a book, there should be a more appropriate font developed for this specific cause, since most cursive fonts are very "handwriting" style. There's also the natural skew of the cursive letters, which looks kind of funny when "straightend up".

Here's a very good attempt in this direction - a font called "Kos Xalav" (Glass of Milk) by Hatayas:

Hey that font is pretty cool. It reminds me of cursive Cyrillic! Funny.
And a[nother] question: it's slanting "backwards" - is this normal?

--

This makes me wonder, in my typical "reformist" way, if there's room to improve the readability of Hebrew by introducing what's basically an "alternative" alphabet (although in a more typographic rendering)... Divergence is good for you, and the conventional Hebrew "print" script is too modular.

hhp

Backslanting Hebrew type is quite common lately in Hebrew fonts, perhaps it's so because of the inverse directionality of the language. In any case, slanted/oblique/italic Hebrew fonts are pretty rare in general because there is no true Hebrew "italic". In an early David Hebrew release (by the late designer Ismar David), an "Italic" style has been suggested, and as you can see it slants to the left, maybe in order to be more compatible with Latin texts (in the link you can also see its beautiful and unreleased "Sans" style I'm currently studying).

It is well known that Hebrew letters lack divergence and therefore Hebrew text suffers great readability issues. There have been some quite bizzare experimentations lately but I still haven't seen a better adaptation of the "cursive" style than the example I showed in my previous post.

> It is well known that Hebrew letters lack divergence

Now if we could just get the Latiners to admit the same about their script...
They're simply too proud.

> bizzare experimentations

Interesting.

--

I think you should make a "cursive" Hebrew book face. I put that in quotes because I don't think it should look handwritten, certainly not obviously so.

hhp

There seems to be much debate about the direction of italic in Hebrew. I had a discussion a few years back with Tzvika Rosenberg on the subject.

David seems to be the only classic font with a true italic and in this case the italic was forward. However, it is my opinion that a Hebrew italic should be backwards in the same direction as its Latin counterpart for the simple reason that Hebrew cursive script (handwriting) is always written in that direction. If italic fonts are supposed to mimic handwriting, then it only makes sense that this should be the direction of the font.

Tzvika likes to call italics that are in the direction of English fonts, "italic" and italics that are forward moving (like in David Italic aka MF Ramot in the Masterfont library) oblique. I personally think that this is an incorrect definition, but the debate will no doubt continue.

Fontbit has just released a new text font with real italics called ktifa, which has italics in both directions in order to keep both camps happy I guess!

> If italic fonts are supposed to mimic handwriting

They're not. They're supposed to differentiate snippets of text in a body of [nominally] upright text.

BTW, I like butterflies, but with that one the urge to swat is palpable. :-/

hhp