FontShrinker - tool to subset a font

[ghostyjack]

I've just gone on your page to download it (2.10) and when I start the install all the text comes up in - I'm guessing here - Dutch.

Is there by any chance an English version available?

[Toxaris]

It should be English actually. I though I had changed that... I will replace it tomorrow. It is a quite normal installer with no hidden or tricky things. It will only install the program and nothing else.

[Toxaris]

Installer changed to the English version.

[Toxaris]

A serious bug is squashed. Also new is a change in the way of working. If from one font family multiple fonts are delivered, the actual usage per font is checked. However, if only one font for that font family is delivered, the usage of the complete font family is checked, not taking bold/italic into account. This is less strict perhaps, but resembles more the actual usage of fonts in ePUB I guess.

[roger64]

Hi

I have two questions

1. - Did you do some statistics about FontShrinker? An average novel uses between 110 to 140 different characters (including spaces, accents, etc.). What would be the average size for a ttf subset using say, Charis Sil? Or could you give an average range?

2. - The editor of Calibre can now provide a csv export list of used characters (though it does not tell to which font they belong). Would such a csv file be useful for FontShrinker?

[Toxaris]

Originally Posted by roger64

Hi

I have two questions

1. - Did you do some statistics about FontShrinker? An average novel uses between 110 to 140 different characters (including spaces, accents, etc.). What would be the average size for a ttf subset using say, Charis Sil? Or could you give an average range?

2. - The editor of Calibre can now provide a csv export list of used characters (though it does not tell to which font they belong). Would such a csv file be useful for FontShrinker?

At 1. No, I test several books and fonts, but the actual reduction is depending on how rich the font was to start with. So the statistics would not really be useful. Now, in the case of Charis, the effect is large, since Charis is a very richt font. So, a lot to remove from it.

At 2. You might also want to take a look at the ePUBOptimizer plugin for Sigil. That is this tool combined with image optimization. Anyway, that list can of course be used. Just paste the actual characters into the box. Be aware that some characters will be added (if they are not yet in the list) like ligatures. The reason is that some readers/reading applications will automatically use a ligature in some cases. So, they are not in the list then but are used.

[roger64]

Thanks for your reply.

1. It's my mistake. I should not have quoted Charis Sil. In fact, I thought that the resulting average size of a subsetted ttf font could be estimated, and to be not so different between fonts. Of course to achieve such a result, the original Charis Sil would have to be downsized a lot while another font would be much less downsized. I was speaking about the end result size.

Another thing that you noted is that there is usually quite a discrepancy between Regular (widely used), Italic (used), Bold and Bold-Italic (scarcely used). This should appear in the end results.

So, I will say it in another way: to provide a book with a Regular and Italic ttf subset for 120 characters, we should embed on average : 120k, 240k, 480k?

2. I'll have a look at EPUB Optimizer.

Nota: FYI these two programs apparently on docs.google.com cannot be downloaded from mainland China. Of course, where is a will...

[Toxaris]

Again, it heavily depends. Let me give a recent example. To a book I added the whole Sabon Next LT font family. That is a regular, italic, bold and bolditalic. Font sizes are (in same order) 146 KB, 142 KB, 129 KB, 123 KB. After optimization the bold and bolditalic are removed (they are not used). The regular is now 61KB and the italic 49KB. The characters used of course differs between regular and italic.
This is just one example. Other one is a font only used for dropcap and headers, Smythe. Standard font is 89KB, after optimization in a book it is 15KB. In the same book is also the Sabon font. The regular size there is 59KB and the italic 40KB.

It really differs per font and per usage.

[roger64]

Thanks for your interesting examples and explanations.

Time now to try it. I hope to be able to add some other figures soon...

[Toxaris]

After a long while, I have made an update. There were some issues that prevented correct behavior. There is now also an option to prevent automated adding of the ligatures to the subset.

[Hitch]

Tox:

A question. Due to the joys of Edge, we are running into a situation that honestly, I hadn't anticipated, with all the font subsetting that we do. The Edge browser overrides any existing formatting that you use, for an ePUB, so, if you have set a particular book to be NOT hyphenated, well, tough.

So, here's a little thing--if you've subset a body font, on an ePUB that does not utilize hyphenation, guess what? The shy is not subset with the book. And, when Edge hyphenates it, lo...you end up with the dreaded Xs.

My question is: when we've chosen, for various reasons, to turn hyphenation off, in ePUB XXX, does the shy get subset, with FontShrinker--or no? I know that you tweaked it to retain ligatures (thank you, Sweetie!), but...the shy? Or do I need to turn hyphenation on, play with the eePUB, allow it to hyphenate, and THEN run FS?

Hitch
(Who feels QUITE cranky about Edge, this morning....)

[Toxaris]

Quite frankly Hitch, I don't know. I always remove the & shy;* myself, so I haven't tested it. I don't think I actively add it to the set to be retained, so if it is not present, it would get deleted right now I think. I will take a look at it when I a back home next week.

[Hitch]

Originally Posted by Toxaris

Quite frankly Hitch, I don't know. I always remove the & shy;* myself, so I haven't tested it. I don't think I actively add it to the set to be retained, so if it is not present, it would get deleted right now I think. I will take a look at it when I a back home next week.

Well, brother:

EEven ignoring me, you're going to want to change that, because good old Edge apparently doesn't know how to replace the & shy;* when it overrides your "hyphenation: off" coding. It ain't pretty!!

Are you on vacay?

Hitch

[JSWolf]

Have you tried font subsetting using Calibre?

[Hitch]

Originally Posted by JSWolf

Have you tried font subsetting using Calibre?

Jon:

Yes, of course we have, and it does not automatically subset the shy, or any other punctuation symbols. OR ligs, for that matter...I think it's a super tool for people to save space, in their own files, but it has some drawbacks (like the ligatures), that make it a better choice, for us, to use Tox's program.

Hitch