Typophile - Spacing method - Comments

Miguel, the first thing that

AGL — Tue, 13 May 2008 03:57:36 -0700

Miguel, the first thing that come to my mind with Calouste is that museum near Spain Square in Lisbon :-) I love it. And adhesiontext is a very good tool. I am sure I will learn something from your sample pdf sample of Calouste. Awesome.

If anyone is interested, I

dezcom — Sun, 30 Oct 2005 09:17:51 -0800

If anyone is interested, I just uploaded a text file in my Blog area which has strings (necklaces) and assorted other pairings that I use to test spacing and kerning.
Here is the link:
http://typophile.com/node/15898

The link is at the top where it makes no sense. The explanation is at the bottom of the thread.

ChrisL

Yes that makes sense.

Randy — Wed, 26 Oct 2005 05:15:03 -0700

Yes that makes sense. Thanks.

“David Berlow: HnHing and

dberlow — Wed, 26 Oct 2005 05:07:05 -0700

“David Berlow: HnHing and HOLAing are new verbs for me. Would you elaborate on these steps? Thanks!”

I use large “unkernable” pairs, like “rv” and “LA” to make sure the kernables, like “AV” and “ra” do not end up “overkerned.” HnHing is to make sure, after the HOD and sometimes V are spaced, that the “n”fits nice to them, so then it can be the prime control for the l.c...clear?

Have you tried adhesiontext?

Randy — Mon, 24 Oct 2005 19:18:54 -0700

Have you tried adhesiontext? Does it address your idea?
Duh. Note to self: Quit skipping over wiki links. Nothing like brainstorming a solution to a solved problem!

It would be pretty easy to do if one had access to a multilanguage open source dictionary database.
Note to self: Quit trivializing said solution when it probably took a great deal of effort.

Thanks Miguel, that is a really useful tool. Not my most graceful moment on typophile :-)

Randy: Congratulations!

Miguel Sousa — Mon, 24 Oct 2005 11:14:43 -0700

Randy: Congratulations! [...] A nice tool would be a database of words in multiple languages with extensive sorting.

Thank you! Have you tried adhesiontext? Does it address your idea?

Hrant: can benefit the user in other ways… like starting on a new design!

Or using another font... :^/

Most of my users (and

hrant — Mon, 24 Oct 2005 10:42:03 -0700

Most of my users (and probably most every other designer’s) are on Word too. But I really don’t see them minding slightly funny spacing. And I will note again that design isn’t about perfection; the time one saves on not making his spacing “always perfect” can benefit the user in other ways... like starting on a new design!

> relying on kerning and adding superfluous pairs to the font

Actually, what I’m describing ends up with fewer pairs to arrive at like 95% of the quality.

hhp

Miguel: Congratulations! I

Randy — Mon, 24 Oct 2005 09:18:14 -0700

Miguel: Congratulations! I came and sat in on an early critique when I was visiting Reading. The quality of your final project is a testimony to your hard work and the excellent program Gerry and crew have crafted at Reading.

All: A nice tool would be a database of words in multiple languages with extensive sorting. So you generate strings containing select groups of letters in select languages. Perhaps this exists already? It would be pretty easy to do if one had access to a multilanguage open source dictionary database.

David Berlow: HnHing and HOLAing are new verbs for me. Would you elaborate on these steps? Thanks!

Randy

I agree with John. Although

Miguel Sousa — Mon, 24 Oct 2005 04:16:25 -0700

I agree with John. Although MS Word users don’t get everything out of their fonts (because kerning is not on by default), they will still get the most out of it, given that the font is well spaced. Glyph sidebearing values have to exist in the font, either we want it or not. Kerning information doesn’t. So, relying on kerning and adding superfluous pairs to the font (and I’m not refering to the extra kerning data that John included for the Ethiopic font, for example), simply makes its size bigger, and adds more data that the applications have to deal with (which means it’s more prone to errors and will use more resources).
My view is, if you can do it with spacing, don’t do it with kerning.

Hrant: The main reason for

John Hudson — Sun, 23 Oct 2005 23:29:34 -0700

Hrant: The main reason for this is that I feel that users who don’t have kerning enabled aren’t picky anyway; so if you can assume that kerning is available, you can streamline things a fair amount.

Given my market, I have to assume that large numbers of users are working with MS Word. Kerning is off by default (although not for complex scripts, thankfully), and most users don’t know how to turn it on. I think such users may well be ’picky’, but if they see a font that has bad spacing because kerning isn’t on, they’re not going to spend time finding out why this problem exists and how to turn on kerning: they’re just going to blame the font and use a different one.

Miquel, my approach to

John Hudson — Sun, 23 Oct 2005 23:25:24 -0700

Miquel, my approach to kerning is to use a list of real words that show combinations in context, so I start with a word that include an ’aa’ combination, and then one with an ’ab’ combination, and so on. I have good lists for Latin and Greek, and a slightly less good one for Cyrillic (it is Russian-centric, and isn’t reliable for other languages). Once I’ve done this, I extend to diacritic letters either using classes or simply by copying kerning, depending on the size of the project. After that, I do diacritic exceptions e.g. ’Tä’ as distinct from ’Ta’.

If I’m working on a script for which word lists are difficult to obtain, e.g. Ethiopic, I use class-based kerning and make it as thorough as possible, even if this means including kern data for combinations that may never occur in natural language. I’d rather try to ensure that any combination will look good, rather than trying to guess what might be common or not. Nyala, which will ship with Windows Vista contains the class-based equivalent of more than 65,000 kern pairs. I usually enjoy kerning, but I didn’t enjoy that! If I don’t have an actual word list, I create FontLab preview lists that show pairs in context of regularly spacing glyphs.

Because almost all of my projects involve OTL tables developed in VOLT, I use Sergey Malkin’s excellent ’kern2volt’ conversion tool, which examines expanded kern table data and builds efficient classes based on kerning values. So my workflow is FontLab class kerning -> expanded (flat) kern table values* -> VOLT via kern2volt.

* FontLab Studio 5 includes special handling for very large kern table export as a result of one of my extended Arabic projects that had more than 72,000 expanded kern pairs.

Miguel, the thing with me is

hrant — Sun, 23 Oct 2005 21:45:24 -0700

Miguel, the thing with me is that I -generally- oppose the conventional wisdom of making the base spacing work “perfectly” without kerning, and only using kerning to polish things up. The main reason for this is that I feel that users who don’t have kerning enabled aren’t picky anyway; so if you can assume that kerning is available, you can streamline things a fair amount. Another reason is actually related to Armenian type design, where relying heavily on kerning, especially positive kerning, can save you a ton of effort - noting here that (to me) in Design, economy of effort is central - and the time I save NOT making “fool-proof” kernless spacing is time I can use for things that are needed more.

Anyway, in reference to what I wrote before what it boils down to is that when my black bodies start getting affected because of spacing considerations, sometimes this is “damped” by the eventual presence of kerning too, depending on the character and its probable adjacencies*. But it’s not numeric, it’s more... fudgy, although it’s also iterative (of course) so subject to tweaking.

* http://typophile.com/node/5106

hhp

“And how do you then

dezcom — Sun, 23 Oct 2005 13:49:19 -0700

“And how do you then translate those values (sidebearings) to the rest of the alphabet?”

With those, I have the sidebearing for every permutation of straight, curve, diagonal, and semi straight (left side of u, right side of h, etc). I only have the left ide of the a and the s and x. They are all between two of the given values (n b o v).

ChrisL

Chris: I then reverse the

Miguel Sousa — Sun, 23 Oct 2005 12:47:47 -0700

Chris: I then reverse the n’s and u’s (unounbunvun) to see if it still holds up.

—And how do you then translate those values (sidebearings) to the rest of the alphabet?

John: and tackle kerning from scratch in a systematic way

—Do you mind revealing?

Hrant: When kerning affects my glyphs, it’s not in any qualitiative way, it’s more of an “abstract nudge” - if that makes any sense…

—Not to me... can you be more specific?... please.

When kerning affects my

hrant — Sun, 23 Oct 2005 10:05:04 -0700

When kerning affects my glyphs, it’s not in any qualitiative way,
it’s more of an “abstract nudge” - if that makes any sense...

hhp

Spacing method

Miguel Sousa — Sat, 22 Oct 2005 15:07:11 -0700

I would like to share with all of you, the spacing method I used on Calouste (PDF file), the typeface I designed at Reading.

After trying out a few spacing methods with no success, I felt the need to devise my own system, which has proven to generate fairly good results (not just for me but also for other people that have used it). The method is definitely not an outstanding innovation, nevertheless I haven’t come across anything exactly like it elsewhere. Basically I took advantage of adhesiontext, the text tool I had developed earlier.

I can’t guarantee this method will work on every typeface. However I’m able to say that it made miracles on mine, as some could attest. Perhaps with a few adaptations it can still be useful and applicable in other cases.

I would also want to warn that, as it will be obvious, this is a REALLY tedious process. Nonetheless, the outcome will most likely be quite rewarding, and time-saving.

Finally I would like to add that this procedure might also help to drastically reduce the number of kerning pairs needed (at least that’s what happened to me...).

The Method Used

1. I divided the lowercase alphabet in 3 groups, bdhilmnopqu, acefjkrt and gsvwxyz, on the assumption that:

a) the space on both sides of each element of the 1st group can be determined, and relates to the space attributed to, at least, one side of another element in the same group, which has a similar shape. (e.g. the space on the right side of b and p is the same as o. Or, the space on the left side of h is the same given to the same side of l);

b) the space on one side of each element of the 2nd group is determined, and is equal to the spaces given to the elements of the 1st group with a similar shape (i.e. the space on the left side of k is the same given to the same side of h, despite the fact that the space on the right side (of k) has no relation with any character in the 1st group);

c) the space around the elements of the 3rd group has no direct relation with any of the other characters in the alphabet. (this is very design-dependent, as for example if g is not binocular-style, it should be included in the previous groups).

2. I spaced n and o visually, balancing the white area between and inside the characters (the string noonnon is good for that). Then replicated those amounts (with the necessary adjustments) to the rest of the elements on the 1st group, always keeping in mind to assign the same value to similar/equal shapes. (e.g. the left side of m might receive the same amount of the same side of n. But (for a serifed typeface) p might need less space on the left side, because it might not have a serif on the baseline)

3. I printed a few test documents only with words that could be set using characters from 1st group, and made the necessary corrections/adjustments.
Note: Here I would like to say that, in this first trial, it was clear that some of the glyphs had to be slightly changed (specially the o), because they were “breaking the rhythm/pattern”.

4. After a *few* other print-outs and tons of adjustments, the elements of the 1st group were finally acceptably spaced, so it was time to do the next step.

5. The next step consisted on adding, sequentially, each element of the 2nd group to the 1st group, and space it against the elements of the 1st group. This meant generating test documents with the set of characters bdhilmnopqua and space the a, then bdhilmnopquc and space the c, and so on.
Note: It’s very likely that the elements of the 1st group will still need some fine adjustments along the way, which could also be used on characters with related shapes.

6. The following step was simply doing the process described above, but this time with each element of the 3rd group.

7. Finally, I set a “normal” text and the overall spacing was just about right. Some fine-tuning, and from this point on was “kerning time”...

That’s about it.