Some notes about rounded fonts

Why manually round, why correct the automatic.
A while ago I have been asked to design rounded versions of a dozen well-known typefaces. I had kept notes and thoughts about the process, in case someone was interested. Here they are.
The tools available
- Manual rounding:
You keep full control of the shapes, but it takes time. The results can be inconsistent. And again, it takes time.
- Christian Robertson’s round script
A good inbetween. Free, reliable, tweakable, but still takes time.
www.betatype.com
- Adobe Illustrator’s rounding options
“Grid too precise”, making designs uneven. Tedious back-and-forth process from font editor to vector editor.
Yotam mentionned this script for AI, also: Round Any Corner
- FontLab Studio 5’s built-in rounding
For some reason, I never managed to make it work reliably, without having FLS move the nodes away.
- RoundingUFO
My solution of choice. A great application, powerful and subtle. Instant support from the developper Frederik Berlaen. Has batch processing, import / export settings,
There could be a more functionnal keyboard input. You still need to watch for slight nodes shifts (±1em)
roundingufo.typemytype.com
- Glyphs
Glyphs has a rouding function built-in, and adjusts the rounding radius depending on the corner angle. But it was not available at the time of this project.
The process
In terms of design, I used some notes from Erik Spiekermann as guidelines for the rounding method, along with FF Unit rounded as a visual hint. The global approach is “the bolder, the mathematically more rounded AND simultaneously the flatness of the terminals is increasing.” Which means that in lighter weights, the rounding radius is less, but there are also almost no flat bottoms.
Another way to put it would be “the blacker, the mathematically rounder; but proportionnaly less rounded”.

Light with a rounding factor of 30, Bold with a rounding factor of 50.
With some exceptions: I cared & corrected, of course, diagonal strokes (e.g. /a/c/e/C/S etc.). I also corrected the rounding amount in smaller glyphs, like superiors. Specific glyphs with thin horizontal strokes (f, t…) have also been tweaked to avoid the “Frankfurter sausage effect”. I also checked angles’ assymetry. For instance, open angles appear rounder than closed angles so I had to increase them. (examples: V W X Y).

Diagonal strokes corrected (left).
It seems that all these tweaks were not all used in FF Unit. But it really makes a difference in terms of consistency: Rounding UFO does a great job in placing the correct nodes, the correct rounding amount, interpreting exceptions in a brilliant way and leaving the rest untouched, but like any machine, it is as dumb as its user is. Those corrections felt necessary to me.

Specific glyphs with thin horizontals need tweaking.
Optically, there are two illusions one might want to avoid: technically they are called “the frankfurter sausage effect”, and “the bone effect” (I think Akira Kobayashi Peter Karow1 first coined the second expression). If you round too much, you can loose the inner tension of the shape, and turn your design into a big, fat sausage. If you do not round enough, an optical distortion can appear at strokes’ ends, making them look like a cartooned bone.

The visual bone effect, exaggerated (left).
Last, I could advice of putting the original font in the background to check it against the rounded version. Another thing: sometimes, the grid is too coarse and putting a point at the rounded top can drastically change your outline. This is a well-known issue to designers. Missing extremes are now most looked at for being responsible for hinting issues. RoundingUFO does not put them on rounded corners and I most of the time I didn’t put them manually either.
1 Update, 2011/02/26: Thanks to Tim Ahrens for correcting the “bone” author, and for mentionning the rounding functions in Glyphs.