I’ve recently rewritten the entirety of a project I did back in October, which combined facial recognition and type design. Most of the effort went into building a new system for how the letters are drawn. Or rather, I should say that most of the effort is also going into learning in great detail more about type design. In the initial version, the letters are drawn with lines and arcs. This was a relatively quick and easy method, considering I made the entire program in about a week, and the results were reasonably effective. However, the limitations of a geometric type system quickly became apparent. (There’s a great article about this here: Making Geometric Type Work )
The tricky thing about making generative typography is being able to allow for a great amount of variation while also maintaining a general level of quality in the letters. There are two notable examples that come to mind: Laikafont and Genotype. LAIKA is drawn with mostly solid fill rectangles, which allows for very nice transitions between weights, serif/sans-serif, line contrast, and slanting. However, this method works a lot better for capital letters than lowercase. Genotype uses bezier curves from common pre-existing typefaces, which gives an almost infinite amount of variation, but also creates a lot of unpredictable results. Fortunately in this case, the unpredictability fit well in the project’s concept.
I made my life really difficult by choosing to draw lowercase letters. And, I chose to attempt italics. This latest revision of the project has significantly better letters than the initial version, but there are still a ton of things I need to fix, and I intend to add pen stroking.
Other than the drawing of the letters, I also added mouth recognition and remapped how all the variables from the face correspond to the variables in the type. Originally, TYPEFACE was about creating a typeface that corresponds to each individual, like a typographic portrait. However, I found that what people really wanted to do was make funny faces at the camera so they could watch the type change. So, I decided to encourage that. I also wanted to draw a closer connection to handwriting, so I added the ability to type in the program, which creates sentences with accumulated variation in the letters.
Next revision coming in March, probably.
edit: Seems like a lot of people would like to download the software to try it themselves. I would like to work a bit more on this project before making a version available for download. Please have a bit of patience, I hope to get this done soon. Thank you for your interest. :)