Join 3,523 readers in helping fund MetaFilter (Hide)


Universal Typeface Experiment
July 10, 2014 7:36 AM   Subscribe

Averaging the world's handwriting to create a universal typeface. Bic is collecting handwriting samples of the Roman alphabet. You can contribute a sample on your mobile device.
posted by ChuckRamone (31 comments total) 16 users marked this as a favorite

 
Wow, that's awesome! Can't wait to see the finished product.
posted by ThePinkSuperhero at 7:51 AM on July 10


Very cool project; thanks for this!
posted by GrammarMoses at 7:52 AM on July 10


Neat! Wish they did lower case, too.
posted by Westringia F. at 7:55 AM on July 10


The USSR is in the list of "where do you live" countries?

(I'm reminded of the time in 2012 when the Air France clerk at Dulles informed me that my bags were "checked through to the Soviet Union" after glancing at the SVO on my luggage tag. We both sort of stared at each other for a moment, a bit puzzled, before she corrected herself and said "Sheremetyevo! Moscow! Russia.")
posted by Westringia F. at 7:59 AM on July 10 [8 favorites]


What has the Roman alphabet ever done for us?!
posted by Celsius1414 at 8:02 AM on July 10 [1 favorite]


Interesting!

I wonder if this would look different if they analyzed actual writing samples instead of having people fill in a single letter at a time. My block printing doesn't really resemble my actual writing, which is generally a mix of block and cursive characters, trending in towards one or the other depending on what I'm writing.

(This is not meant as a criticism of the project. Just a further interest.)
posted by jacquilynne at 8:02 AM on July 10 [7 favorites]


I wonder if this would look different if they analyzed actual writing samples

Wish they did lower case, too.


What has the Roman alphabet ever done for us?!


I'm going to hazard some guesses — they want to keep this as simple as possible. So lower case is 26 more characters to ask people to contribute, which could lower participation. Going beyond the Roman alphabet, same problem, many more character sets. Actual handwriting samples: also more complicated to carry out, but more importantly: a typeface is not handwriting (so there's some problematic terminology in how they frame the experiment). This is really an experiment in collective design of a simple font, not an experiment in averaging handwriting. My guess is that in the end, they are going to produce a font out of those heatmaps by drawing hard edges to include just the portions that exceed a specific density.

(And another guess: the result will be Comic Sans.)
posted by beagle at 8:16 AM on July 10 [8 favorites]


I can't see that URL as anything other than The Univer Salty Peface for some reason.
posted by Rock Steady at 8:18 AM on July 10 [7 favorites]


There is no possible benefit that could come to this project from my submitting a sample of my handwriting. Cool project though!
posted by obfuscation at 8:24 AM on July 10


Actual handwriting samples: also more complicated to carry out, but more importantly: a typeface is not handwriting

From my basic understanding of typefaces, they are idealized handwriting. So, to average a bunch of people's handwriting and then make some kind of composite typeface from it could be considered an idealized handwritten script. Then again, average could be considered shit as well, depending on whom you ask.
posted by ChuckRamone at 8:31 AM on July 10 [1 favorite]


Collecting that many samples would also be useful for handwriting recognition. Unfortunately, by prompting the users to write using a specific typeface, they're biasing the results. I'm not really sure how else to do it though. Randomize the prompt typeface?
posted by pwnguin at 8:50 AM on July 10


If they add my handwriting to the mix, the typeface will most certainly be called 'illegible'.
posted by Pudhoho at 8:55 AM on July 10 [2 favorites]


I think the letters that don't close in the averaged version are interesting. For example, Q doesn't close. I assume the algorithm that does the averaging makes some sort of smoothed curve from the points along each person's letter. But it seems like it should also pull in other properties of each letter somehow -- this stick tends towards the vertical (but how vertical would be averaged), and these two lines tend to touch (though where would be based on the average). Otherwise it seems like the idealized letters could have properties (like non-closing) that are in a minority of the sample letters.
posted by If only I had a penguin... at 8:57 AM on July 10


Needs more southpaw.
posted by Buttons Bellbottom at 9:13 AM on July 10


I wonder if this would look different if they analyzed actual writing samples...

There's a word of difference between the letterforms people can scrawl on a screen and what they scrawl on a piece of paper. That this experiment is limited to a digital interface is disappointing, imho. It's more an averaging of screen gestures, rather than handwriting.
posted by Thorzdad at 9:14 AM on July 10 [3 favorites]


pwnguin - audio prompts?
posted by Westringia F. at 9:15 AM on July 10


Needs more southpaw.

Here's one more in. If only this could capture the extent of smudging that my writing develops whenever I use a slow-drying ink or a pencil.
posted by pemberkins at 9:27 AM on July 10


Collecting that many samples would also be useful for handwriting recognition. Unfortunately, by prompting the users to write using a specific typeface, they're biasing the results.

I wondered about this, too. I wrote "I" without crossbars, and then couldn't figure out if that's how I normally write "I" or if I just did that because that's how it was presented to me. Looking at samples on my desk, it appears that I write "I" with crossbars when filling out forms / deliberately block printing, but without when writing my usual casual mix of cursive and printing. So probably I would normally have used cross-bars in this context, even though for the most part I don't.
posted by jacquilynne at 9:28 AM on July 10


Needs more southpaw.

You can compare left handed and right handed averages of the individual letters. The ones I looked at all seemed pretty similar. I was struck that there were occasionally marked differences in letter form by gender, which seemed surprising (check out the W for example).

One thing though--when I tried to look at results by country, I couldn't seem to find anyway to scroll down--I would just get stuck with the top six or so countries in their list. Anyone figure that out?
posted by yoink at 9:29 AM on July 10


Previously: Averia font, which is made form an averaging of typed fonts. Linked linked article gets into a ton of nitty gritty behind the scenes it's-actually-more-complecated-and-awesome-than-you'd-expect stuff
posted by rebent at 9:50 AM on July 10 [2 favorites]


(I used Averia on my wedding invitations!)
posted by rebent at 9:51 AM on July 10


Looking at the Y, the vast majority of the people make a slanted Y, like the lowercase y in verdana but bigger, rather than the upper-case Y in verdana that meets at 3 points. The yellow line still shows the 3-point version, though, which is making me wonder if the yellow lines are really the averages, or if they're just illustrations.
posted by zug at 9:57 AM on July 10


There's something ironic about a pen brand collecting handwriting samples for a digital typeface.
posted by monospace at 10:06 AM on July 10


"There's a word of difference between the letterforms people can scrawl on a screen and what they scrawl on a piece of paper."

It seems to me that they're a bit premature for this. I think the mouse-created samples should be considered invalid at the outset. But touchscreen-created samples, and especially those using styluses, would be great. We're very close to a world where the latter is more prevalent than the former; they'd only need to wait a few more years to get primarily touchscreen-created data.
posted by Ivan Fyodorovich at 10:19 AM on July 10


I have an image of various distressed, obsessive font geeks around the world desperately making repeated, exaggerated contributions in hopes of correcting various letters that are turning out all wrong.
posted by straight at 10:49 AM on July 10 [4 favorites]


Unfortunately, by prompting the users to write using a specific typeface, they're biasing the results.

The obvious solution would be to use lower-case letters as the prompt, and off to the side so aren't used as patterns.

As it is, it seems more like an exercise in averaging people's ability to trace accurately with a mouse.
posted by straight at 10:53 AM on July 10 [1 favorite]


pwnguin - audio prompts?

App: Ready to begin? Print a capitalized A.

Arthur Fonzarelli: AAAAAAAA ....
posted by Buttons Bellbottom at 10:58 AM on July 10


can't see that URL as anything other than The Univer Salty Peface for some reason.

The kerning on Univer Salty is atrocious.
posted by sidereal at 12:24 PM on July 10 [4 favorites]


I really would like to like this. Is there some part of the website that I haven't been able to find that does away with all the Web 2.0 (or whatever web we're currently on) crap and just allows me to look at a picture of the overall average of all 26 letters at once, instead of one (or three or whatever) gender-or-age-segregated letters at a time with things scrolling around of their own accord and random people's specific contributions overwriting things that I'm looking at?
posted by Flunkie at 12:42 PM on July 10 [2 favorites]


Hmm...I was hoping for something along the lines of "Write the letters on paper and take a picture of it for scanning". My fat fingering on my phone looks very little like my actual pen-based handwriting.
posted by Mr. Big Business at 1:09 PM on July 10


beagle, I was thinking "comic sans" as I read your suggested methodology. I was not disappointed to see you reach the same conclusion, it feels almost inevitable
posted by aydeejones at 4:39 PM on July 10


« Older Since 2010, Breanne Fahs, associate professor of w...  |  Billy Hanshaw is a VFX artist,... Newer »


This thread has been archived and is closed to new comments