Erm, this auto-spaces, right? I don’t think spacing is an after-thought. The way this is presented doesn’t smell good to me—it’s not kerning, and it is advertised badly.
I think what makes this project more attractive than other automated kerning systems is that type designers have a very real chance to make it better. Because Autokern is open and written in Python/Robofab there is a significant pool of potential users who can jump right in and contribute.
Why? We already use interpolate entire fonts out of existing data. If someone can create an algorithm that can take font data and some parameters and crank out good kern tables, why is that a horrible idea?
It can alter glyph spacing, but that can also be disabled. But the success of iKern seems to suggest that with the right person at the controls, a computer can tweak spacing just fine. So maybe having a mechanical third party to critique spacing isn’t so terrible. There are plenty of good typefaces that just have one letter sticking out here and there because a sidebearing is too wide and the designer never noticed.
I think autokerning is a wonderful idea, as long as you review and fine-tune by eye. There's nothing expressive about kerning. It's a matter of mechanically evening out light and dark bits to create an even color. Why not use a machine to accomplish a mechanical process?
I don't agree with that. Spacing (including kerning) is an important part of the design of a typeface. There are different styles and different methods, which can make a big difference in the way a typeface looks. Automated spacing (or kerning) should take this into account, if it can.
Regarding mistakes in spacing and kerning due to it having been done manually, there are Python scripts that can check for things like that.
Hi there, I'm the author of this project. Thanks for taking a look.
Yes, I agree many of the current results are horrible. Autokern is very much a work in process, and I've only been working on it for a 4 weeks. If any of you are interested, I would love to get a more detailed critique.
I'm a software engineer, not a type designer, so unfortunately I do not truly understand the problem domain. Your thoughts would be very welcome. charlesmchen@gmail.com
I'm trying to learn about type design and this project is one of the ways I'm pursuing that. Ideally, it will bring me into contact with other people who know more about the subject matter and from whom I can learn. ...and I enjoy coding anything font-related.
Looks like there's a lot of people who disagree that kerning's essentially mechanical. And, obviously, spacing a typeface is a key part of the design process, and the decisions you make there have a lot to do with the face's spirit and personality. But once you've done the basic spacing—and made the basic decisions about how letters should fit together—it does seem to me that kerning's simply a matter of making the more awkward pairs visually match the rest. And I'm not sure I see the expressive potential in that particular part of the process.
A fair number of well-regarded designers use automated kerning (and spacing, for that matter). Do folks think that iKern's clients, for example, are putting out inferior work? Mark, have you ever looked at a typeface and thought: Too bad the kerning's so mechanical and uninspired?
No, I was just saying I would hope that an automated tool would not make too many assumptions about what constitutes "ideal" spacing and kerning. It's partly a matter of personal preference.
I only notice when the spacing is bad or uneven, too tight, word space too large. Things like that. I'm not sure I would be able to tell the difference between a font that was hand-spaced and one that was auto-spaced. Hand-spacing does not necessarily mean better spacing.
Well, if a trained eye like yours can't see the difference between machine kerning and hand kerning, that suggests to me that kerning (separate from spacing) isn't that expressive a process. That's all I was saying.
For all I know, the fonts with the worst spacing were done automatically. I would be surprised if the fonts with the best spacing were done automatically. Just because I can't tell doesn't mean it's the best way.
The thing I don't understand is the idea that spacing would be a completely separate step. "Finally, done drawing all the characters. Now on to the spacing!" I'm working out the spacing as I draw the letters. I may discover a spacing problem that is caused by the shape of a letter, so I fix that. It's a back and forth process. Kerning may be a bit more successfully automated, but doing it manually you may find better ways to solve spacing issues other than adding kerning pairs.
The thing I don't understand is the idea that spacing would be a completely separate step.
I agree with you at least when it comes to working out the baseline for spacing. I would never let software muck around with control characters, or for that matter, change my sidebearings at all. But I wouldn’t mind having an objective machine “eye” that can analyze what I did with the control characters and then tell me that K and 7 are loose in comparison. This is also how I think auto-kerning could work well—analyze the controls and then even out the rest. But I don’t know if any of the approaches on the market work that way.
The thing I don't understand is the idea that spacing would be a completely separate step. "Finally, done drawing all the characters. Now on to the spacing!"
Of course, type designers are part of the problem here. How many of us tell people who want to design type to go read Letters of Credit, or Designing Type, both of which describe drawing and spacing as different acts? Almost every amateur type design thinks this is how it is done. So they just want to draw in AI, automagically dump it all into Fontlab, space it, and then just be done. Maybe someone with more credibility than me should write an ILT article explaining that this was done in the past out of necessity, and that it no longer makes sense.
Perhaps kerning could count as mechanical process, but good kerning can not.
I object. ;-) What is it, that you could measure with you eye, that a computer cannot? If auto-kerning is bad, than the algorithms are still to bad. If you would put enough effort in it, you can easily match or outmatch human kerning. Soooo many aspects of type design can be broken down to rules and algorithms. The fact that fully parametric type design isn't here yet, is not that humans are intrinsically better in drawing type, but just that no one has yet taken the effort to create a usable tool with enough control parameters to match the current state of "regular" type design.
Maybe someone with more credibility than me should write an ILT article explaining that this was done in the past out of necessity, and that it no longer makes sense.
Comments
What do you think of cars that will parallel park themselves? :-)
I think it takes all the fun out of it.
Regarding mistakes in spacing and kerning due to it having been done manually, there are Python scripts that can check for things like that.
Yes, I agree many of the current results are horrible. Autokern is very much a work in process, and I've only been working on it for a 4 weeks. If any of you are interested, I would love to get a more detailed critique.
I'm a software engineer, not a type designer, so unfortunately I do not truly understand the problem domain. Your thoughts would be very welcome. charlesmchen@gmail.com
A fair number of well-regarded designers use automated kerning (and spacing, for that matter). Do folks think that iKern's clients, for example, are putting out inferior work? Mark, have you ever looked at a typeface and thought: Too bad the kerning's so mechanical and uninspired?
I only notice when the spacing is bad or uneven, too tight, word space too large. Things like that. I'm not sure I would be able to tell the difference between a font that was hand-spaced and one that was auto-spaced. Hand-spacing does not necessarily mean better spacing.
The thing I don't understand is the idea that spacing would be a completely separate step. "Finally, done drawing all the characters. Now on to the spacing!" I'm working out the spacing as I draw the letters. I may discover a spacing problem that is caused by the shape of a letter, so I fix that. It's a back and forth process. Kerning may be a bit more successfully automated, but doing it manually you may find better ways to solve spacing issues other than adding kerning pairs.
Soooo many aspects of type design can be broken down to rules and algorithms. The fact that fully parametric type design isn't here yet, is not that humans are intrinsically better in drawing type, but just that no one has yet taken the effort to create a usable tool with enough control parameters to match the current state of "regular" type design. Sounds good!