Stuck on embedded Hebrew font in ePub

[Gringo]

Hello folks!

A company converted half of a job to ePub format, and when they quit, it was handed to me. Things had been humming along ok until a couple of the books started using a Hebrew font. So I fiercely read up on how to embed fonts. The only problem is that the Hebrew font used to create the original book (in InDesign) is not easy to work with. From my testing, it will not embed. (Here it is: http://cl.ly/1m1e0N1E1k2S) And when I use a different Hebrew font which can be embedded, it shows up like this.

But now I'm confused, as I look over the work they had done previously. For example, when I explode their ePub in Calibre and check out their text, I find:

O verbo hebraico, זגר (rāgăz), significa ou ser agitado pela ira, ou pelo temor.

So that three letter word is good Hebrew, so I looked up their xhtml file:

O verbo hebraico, <span class="simbolo">זגר</span>

And their freaking CSS file has only:

span.simbolo {
	font-family: "Times New Roman";
	font-weight: normal;
	font-style: normal;
}

The heck? How did they get that Hebrew text to show up? They worked off of the Adobe InDesign publisher's file just like I am, and did they simply go copy/paste the Hebrew word directly to the ePub files? That would have taken forever — because Hebrew words are all over the place!

I've run laps around the web looking for a break but haven't found a fix. Any help is appreciated! I know I'm missing something — I just don't know what yet.

[DomesticExtremis]

Micro$oft's TNR font covers the Hebrew Unicode range.

[Doitsu]

Originally Posted by Gringo

O verbo hebraico, זגר (rāgăz), significa ou ser agitado pela ira, ou pelo temor.

First of all, Hebrew is written from right to left. However, in the example sentence that you posted, Hebrew text is written left to right. (זגר reads Z-G-R from right to left; however it should be displayed as רגז = R-G-Z). I.e., before fixing the font issue, you'll have to figure out how to format RTL text in InDesign so that it'll export correctly. If you only have the ePub to work with, you'll probably have to manually rearrange all Hebrew letters unless the the author only used three letter Hebrew roots, in which case you might be able to rearrange them with a regular expression.

Originally Posted by DomesticExtremis

Micro$oft's TNR font covers the Hebrew Unicode range.

Ezra SIL is free and will do nicely.

[Toxaris]

Since on the readers Hebrew is usually not includeded (if any at all). you indeed need to embed the font. To keep the size under control, I advise to subset the font.

[Jellby]

Originally Posted by Doitsu

First of all, Hebrew is written from right to left.

With a Unicode-compliant system, you'd enter the letters in their natural order (the order in which they should be read), and the system should take care of arranging them left-to-right or right-to-left, depending on their "intrinsic" directionality (which is rtl for Hebrew letters, of course).

But ADE does not support right-to-left in any way, manual or automatic, so your only choice would be entering the letters in reverse order, which is wrong, and would show wrongly in compliant systems.

[Doitsu]

Originally Posted by Jellby

But ADE does not support right-to-left in any way, manual or automatic, so your only choice would be entering the letters in reverse order, which is wrong, and would show wrongly in compliant systems.

You're of course right. The only way to display Hebrew correctly on ADE/RMSDK-based readers is to reverse the order of the letters. (Since five Hebrew letters have special final forms, these forms also need to be hard-coded.)

However, since the OP apparently owns an iPad, I was assuming that he wants to create an ePub for the iBookstore. Since iBooks supports RTL text, it'll display זגר instead of רגז.

The only way to display RTL strings correctly on ADE and non-ADE readers is to use your XPGT method to show LTR text on ADE readers and the RTL text on all other readers. I used a variation of it with embedded Arabic text. (The Arabic text was included twice: one pre-shaped LTR version and one regular RTL version.)

[Monica Xiao]

Have you fixed it? I met the same problem. Looking for a solution....

[Doitsu]

Originally Posted by Monica Xiao

Have you fixed it? I met the same problem. Looking for a solution....

AFAIK, the only solution for RTL ePubs that work both with ADE devices and iBooks is to include RTL text twice (RTL and LTR) and hide one of them. Have you looked at my Arabic demo file?

[mrmikel]

One thing that was only alluded to is that Times New Roman is NOT a free font. It is one which probably require an expensive contract with the font provider. That is why Ezra SIL was mentioned and it or something like it needs to be used.

[odedta]

Gringo, when you put hebrew letters in an HTML file within an ePub and load it on your iPad, you will see the Hebrew letters perfectly and when you look at the font it will say "Times New Roman", you don't need to embed a hebrew font in order to see hebrew letters (in the correct order - RTL), at least on the iPad.

As for:

<span class="simbolo">זגר</span>

Doitsu explained it really well, basically it should say רגז and not זגר
It probably happens since your InDesign version does not have hebrew support.

You can still try to use the CSS rule: "direction:rtl" for the .simbolo class, i'm not a 100% sure it will reverse the letters order, if it wouldn't, you either have to manually reverse their order which can be done via Regex or add hebrew support to your InDesign.
If you need to convert just a couple of files I can help you with that.

by the way, the image you posted here looks weird, those symbols between the hebrew characters ס and פ doesn't look familiar, it feels like the font is not a hebrew font and that's why the device finds a different character from the charset

[Hitch]

Originally Posted by Doitsu

AFAIK, the only solution for RTL ePubs that work both with ADE devices and iBooks is to include RTL text twice (RTL and LTR) and hide one of them. Have you looked at my Arabic demo file?

Which is why we turn down jobs that have both English and Hebrew in them. Far, far too much brain-damage. I try really hard not to think about individual pages in which we're supposed to have both, or, yet more fun, facing pages with, say, English (l-to-r) on the verso and Hebrew (r-to-l) on the verso. Thanks, but...no thanks. Certainly not without charging the bloody earth. The conversion house that took that job were idiots if they didn't know upfront that Hebrew was going to be a problem for a professionally-produced book. And let's not forget those page flips!

(This isn't to say we don't do single-language r-to-left; we do. But mixing them up? Oish).

Hitch

[DaleDe]

I dealt with some mixed text in the wiki and I just put the Hebrew letters text and English words on different lines and let the html viewer merge the lines as would be expected. Lo and behold it worked.

Dale

[odedta]

I really don't see the problem there Hitch, dir="rtl" would solve all your problems :P

[Jellby]

Originally Posted by odedta

I really don't see the problem there Hitch, dir="rtl" would solve all your problems :P

Not even that, just write Latin and Hebrew letters normally, the Unicode bidi algorithm deals with the rest

[Hitch]

Originally Posted by Jellby

Not even that, just write Latin and Hebrew letters normally, the Unicode bidi algorithm deals with the rest

Not to be uncommonly dense, but: when they're on the same page, which way do the page flips flip? Oh--I see why you guys are confused. This is a FF book, which means, automated page flipping.

On a more serious note: the books we've turned down aren't one or the other. I am not making this up: one had English on the verso and Hebrew on the recto, and the client wanted it to auto-play, optionally, in both directions. Now, I'm all for fun and games, if someone is paying, but honestly: I'm not sure I'm that interested, commercially, in making a book that flips in both directions, automatically. It sort of makes my brain hurt. I'm opting, at the moment, to let Liz or the guys at Azardi or odedta figure it out first.

And is someone here claiming that it will work in all readers? Methinks not.

Hitch