Interesting. In 2000 years time this may be seen as a quality in that face, like the 'effect' carving in stone leaves on letters.

What are you, the talk show host of the type world? ;-)

> I think that they idea of bundling three separate italic families together is a great one

Thank you.

> Type design isn't about the tools used

True, but I personally don't mind the presense of the tool as long as there's no clear detriment. A much more pervasive -and demonstrably harmful- tool present at the core of Underware's work is the hand. You can see one strong manifestation of this in their new Typographica banner.

Don't get me wrong, Underware makes well-crafted, superbly usable and way-better-than-the-norm fonts (which is very hard to do already), but they're fonts that celebrate, not really push, culture. And to me the younger you are the more that's wasteful of potential.

hhp

>What are you, the talk show host of the type world? ;-)

yes.

>You can see one strong manifestation of this in their new Typographica >banner.

exactly my point! That is so bezier-y in its appearance� you can just see the way the lines curve, how it was drawn, with which tools� at RISD we used to say that things like that were "so about the computer"

I've been thinking about this because I'm working on a light-weight typeface now, and I'm trying not to make the curves, especially italic curves, look so bezier-y. It is really hard.

> yes.

Just don't get your nose broken. ;-)

Concering the Typographica banner, I was actually referring to the way the "com" is made to write out. A font that's writable is not enough of a font. And then look at the Bold "A" of Unibody on that banner: fealty to chirography causes it to be not only distracting, but just plain ugly. If I remember correctly that also happens with its "y".

hhp

If they were to lay text set in Auto overtop of foam core, and then cut that, then yes, it would still look the same (bezier-y)

If they would carve forms directly out of foam core without an overlay, I doubt that they would look exactly like this.

You are certainy right in that they most likely intended to make their forms look this way, I'm sure that they all had complete total technical control of their project from beginning to end. My feeling though is that intentionally left some of their tool in the final product (maybe to show process, or whatever). It is this that perplexes me. Why make your letters look bezier-y? The dangers of this "style" have already been elaborated above by Hrant.

I don't think that their other typefaces look bezier-y.

yes, it is mostly the lighter weights that give me this feeling

> It seems as if they don't bother, or didn't with Auto, to think of the "pen" at all.

?!
It's plastered all over it.
Most notably in italic 3, where those bulges at the bottom-lefts and top-rights of stems are so clearly direction changes of the pen.

hhp

With the exception of that third italic which is too stiff and mannered for my taste, it looks like a nice �contemporary� typeface to me.
I wouldn't mind recommending it the next time I'm asked to do some typographic scouting.

I myself have no use for three italics, but there's all kinds of people, obviously.

�

Dan, thanks to explain what mean "bezier-y"?
I know Mr B�zier, but bezier-y? his son?

What is that new subfoteallinearsanschirobezier-y+z6 concept?

I recently found a photo of the archetypical chirographic designer:



The alternative I recommend? "Think with your brain." :-)

hhp

By bezier-y, I mean that you can just "see" the bezier curves inside the letters' strokes (how they were bent around, where the nodes probably were, the handles, their angle, etc.)

I know that all typefaces are drawn with Bezier curves (except that whole TT spline thing), but normally, when I look at a face (say, Sabon NEXT for example), I just see forms and curves� I might think� oh, I could draw that curve! (be it with a pencil, a pen, digitally, whatever) But with the thin weights of the Auto Italics, I look at the curves, and just think� hmmm, they've had fun making those bezier curves.

I too am intrigued by the effect that the underlying curve primitive has on font design. I think that designing on the screen, as opposed to making drawings and then basing the digital outlines based on those, makes a significant difference as well.

Indeed, I feel so strongly about this that it now seems likely that I'm going to do a PhD on the very topic. I believe a better curve primitive is the Cornu spiral. I've just built a prototype curve editor using the primitive (see screenshot) and so far am very encouraged that the resulting curves are smoother and more natural than those designed interactively with Beziers.

No doubt you're going to hear a lot more on this from me, but I wanted to post a quick note now because it seemed on-topic.

Now that's a PhD topic. Godspeed.

Question:
What kind of translation loss would you expect going from Cornu spiral to Cubic/Quadratic bezier?

hhp

Sorry Hrant, but your image is pretty funny, but doesn't help.

Thanks Dan! I get it now.

Trying to explain what you see:
What I see on that new face, is they perhaps designed from a monoline, them transformed as path (such what you can do in Illsutrator)? Beside that, i just see shapes produced from the usual The Hague philosophy in bit less serious way? the forms rely pretty on good old Chancery script (the image on typographicacom).

> your image is pretty funny, but doesn't help.

Eh, such a functionalist. ;-)

hhp

hrant: that's a complex question. The short answer is that you can go from Cornu spiral to quadratic or cubic Bezier with insignificant loss, as long as you have enough precision in the coordinates [in fact, my interactive editor converts to Beziers before rendering to the screen]. So there's no problem outputting PostScript code that has high-precision float coordinates followed by "curveto" operators, but an unsophisticated transform to TrueType or Type1 can introduce significant errors.

So, in order for this stuff to be useful for type, the output module has to be smart enough to avoid those errors (briefly, very short arcs are the enemy). Of course, the exact same thing is true of stroke offset algorithms (the constant-distance offset curve from a quad bezier is a complex fifth-order polynomial that can get quite numerically snarly where the offset radius gets close to the curvature of the Bezier).

(the constant-distance offset curve from a quad bezier is a complex fifth-order polynomial that can get quite numerically snarly where the offset radius gets close to the curvature of the Bezier)

This tempts me to say something like: My sentiments precisely but then I don't know what I'd be saying. I'll just say I'll bet...

A doctor of Cornu spirals. How cool is that?

As for the Auto series i think some of the glyphs are pretty entertaining (like the ampersands) but I can "see the beziers" too like in the tail of the "et" ampersand especially.

Stare into the bezier...

Jon

i like the vigour & springyness of the auto forms, reminds me of the images in their workshops.

the whole bezier-y thing. surely its all about exposing materials & methods to the passer-by. like a timber frame building or mies' I-beam buildings or somthing; ruskin-stylee. that's also what i got from that beezer sans, posted on the critque section ages ago.

personally, i'd sooner have the auto family than something like brioso.

>the usual The Hague philosophy

J-F, how would you characterize that philosophy?

>a better curve primitive is the Cornu spiral

Very interesting. Why the Cornu spiral as opposed to other spirals? Good luck with the dissertation!

The auto typeface approaches the third italic as a cursive, and to my knowledge it's the first that attempts to buid that into a family. I think we might also be able to charictarize Auto 3 as an upright cursive italic. The stems are verticle, but curves are cursive, and the overall effect is italic.

I think that this innovative and bold design reflects our culture's irreverant attitude, sure, but the fact that we're put off by the strange look of some really different letter forms tells me that they are pushing things, at the very least, our buttons.

Aaron

Why the Cornu spiral as opposed to other spirals? Good luck with the dissertation!

Thanks! That's a good question. I'll try a brief answer.

1. It's one of the few spirals that has an inflection point in it, or in other words contains an "S" curve. Most other spirals only wind one way.

2. In fact, it's got a number of useful shapes hiding in there. At the inflection point (t=0), it closely fits a basic cubic polynomial (essentially the same curve you get when trying to make an "S" in a cubic Bezier). At roughly t=0.6, it closely fits a parabola, which is the geometric primitive of a quadratic Bezier (TrueType). And as t increases, it closely fits a circle.

3. It's the smoothest way to interpolate between curvatures. In fact, designers of roller coasters and highway on/off ramps use the curve to smooth out the lateral g-forces. If you were to ride a motorcycle along, say, the 's' from Auto Italic 3, you'd be in for some rather violent shifts of g-force.

4. It has deeply beautiful mathematical qualities. The more I study the curve, the stronger my intuitive feeling of rightness. For the details, you'll have to wait for my thesis (how's that for a cliffhanger?).

I should have linked a PDF of the Cornu spiral, so that you can see the curve I'm talking about.

Type design does expose materials and methods. To me, cubic Beziers are like carving letterforms into rough pine wood with knots (pun intended, of course), while Cornu spirals are like carving into pear wood. If you're going for an effect, the former makes sense, but given the choice, most of the time a seasoned craftsman will go for the latter.

> it's got a number of useful shapes ....

That trio of "shape-convergences" shall we say does indeed make it seem very promising.

hhp

>At the inflection point (t=0), it closely fits a basic cubic polynomial

I get it. Bezier curves are easier to work with than quadratic splines because, being based on cubic equations they can do the 's' shape, which would take two quadratic splines, and hence more 'knots' to move around. And you're saying the Cornu spiral can do this and a lot more.

I guess the question is whether you can put the equation in some kind of parametric form with 'knots' and 'handles' as Bezier did. Bezier has the two end points and the two handle points that define the curve, which you can move and see how the computer redraws the cubic curve.

It sounds like the Cornu spiral is more complex than a cubic equation. Will you need a fifth point, a third 'handle', to adjust other features of the 's'? So then your work would look something like a Bezier and a quadratic spline together - an off-curve point, and the two tangent handles? Or will you need a fourth handle as well? There is the danger that it will be too complicated to work with.

Sounds like a great idea - best of luck with it!

It sounds like the Cornu spiral is more complex than a cubic equation.

It depends on what you mean by "complex". Beziers can be computed using only high school algebra, while the mathematical basis of the Cornu spiral is the Fresnel integral, a "special function" that requires a bit more mathematical skill to work with.

But if you're talking about complexity from the user's point of view, Cornu splines win hands down. Once you nail down the endpoints, a Cornu spiral segment has only two degrees of freedom, the same as a quadratic Bezier (and two less than a cubic). In short, it's as simple to use as a quadratic, but every bit as powerful as a cubic.

Of course, the real test is how natural it all feels in the context of an interactive editor. I've built one (for Linux only, so far), and so far am very encouraged. I'm eager to get it in the hands of typophiles (as opposed to math nerds) so I can get good feedback from the real world.

Thanks again for the encouragement! I have a meeting with my advisor tomorrow, so all this feels like a wind at my back.

>Cornu spiral segment has only two degrees of freedom ...two less than a cubic.

But isn't it the *two* handles (each moving in two dimensions) that make the Bezier curves easy to work with, compared with quadric splines?

Maybe the Cornu spiral will seem so natural that one will get a good result with only the one point in addition to the end points, but I wonder.

Thanks for the link to the cornu spiral picture, Ralph. Your project is fascinating. I presume you're familiar with the work of Philippe Coueignoux, whose parametric font design system (never more than experimental, I think) he made while at MIT in the 1970s used spirals of some kind.

Can you elaborate how you can transform spiral segments? Do you use 3D perspective transformations, for example?

I love your comment about "designers of roller coasters and highway on/off ramps". I've occasionally wondered if there's any maths in the trajectory that the good driver tends towards, when attemping to drive quickly and (therefore) smoothly.

Laurence: yes, I have come across the idea that Coueignoux used spirals, but have had little success tracking it down. His 1975 thesis seems to be based on polynomials.

Is my memory faulty, or did he not go on to work for one of the major typesetter vendors of the day? Possibly Compugraphic? Perhaps there was a later proprietary font format that was based on spirals.

A little more background on Coeignoux is in this note by William Schreiber.

> Once you nail down the endpoints, a Cornu spiral
> segment has only two degrees of freedom

And with slope continuity (with adjacent segments) at those endpoints?

hhp

And with slope continuity (with adjacent segments) at those endpoints?

Right. Enforcing the tangent continuity across a knot point eats up one degree of freedom. The best representation, then, for a manually drawn Cornu spline is a "spinner" for the tangent angle at each knot point. Using an off-curve point instead of the tangent angle has the drawback of not handling "S" curves well; the two tangents could well be parallel to each other, so they would never intersect.

The curve editor I'm working on assigns the tangent angles automatically, as well, so all the user needs to put in are the on-curve points. I'm going to make the tangent spinner optional at each knot; it seems to me especially useful at corner joins and joins with straight line segments.

At knots where the tangent is chosen automatically, there is curvature continuity as well as tangent continuity. My sense is that when typophiles talk about the "bezier-y" feel of an outline, it corresponds fairly closely to the mathematical concept of disconuities (or wide variations) in curvature. Cornu splines are extremely good with respect to that criterion. Perhaps even optimum; it's a question I want to study.

Maybe the Cornu spiral will seem so natural that one will get a good result with only the one point in addition to the end points, but I wonder.

That is more or less my claim. Since the concepts of "seeming natural" and "good result" are so subjective, I intend to test that claim by putting the curve editor into people's hands.

Can you elaborate how you can transform spiral segments? Do you use 3D perspective transformations, for example?

No, it's simpler than that. I just cut a piece of the original spiral (the two degrees of freedom mentioned above correspond exactly to the points along the original curve where I cut), then scale, rotate and translate so that the endpoints of the spiral section coincide with the desired endpoints.

> putting the curve editor into people's hands.

Cool. If you manage an online Flash thing that would be killer.

One last question (for now): what about straight lines?

hhp

> a "spinner" for the tangent angle at each knot point

Neat idea, sounds good. I see on the links in your profile that you have patents. I hope you are protecting your idea legally, as you deserve to benefit from it if it works out.

It sounds like this would be a great addition to FontLab, where you could have a choice of working in the Cornu spiral, as well as the existing cubic beziers, quadradic splines, and the 'sketch' mode.