WhatTheFont vs. IdentifyFonts -- what's the better service?

WhatTheFont is pretty much the established player in the font identification market. The service's claim to fame is its ability to figure out what font is used in a particular image (or try to, at least).

A new contender, called IdentifyFonts (original naming there!), recently came into this same space. Since they are so similar, I felt a head-to-head comparison would be the obvious thing to do.

First of all, I must admit that I tested both services with a challenging image. It's white text on reddish background, and the typeface is a thin sans-serif. So, neither of the services had an easy task, but one definitely came out on top. To see which service took the round, keep on reading after the fold.

IdentifyFonts

Once you upload an image (or provide an image URL), IdentifyFonts lets you step through a very long screen where your image is displayed over and over again, with a different character blacked out every time.

At each step, you're supposed to type the character you see; you do this for each and every character on the image, so it's a pretty time-consuming process.

This is where things went horribly wrong for this fledgling service: it simply and completely choked on my image. As you can see in the screenshot above, it blacked out the "hole" in the middle of the character and not the character itself.

At first, I thought it was just a rendering quirk and that was how they marked the characters they wanted me to enter. Dutifully, I went through with the process.

When the results screen came, the truth became apparent; my results had nothing to do with the delicate sans-serif used in the image. The service threw a bunch of blocky, hole-ridden, "artsy" fonts at me. I guess it really thought the holes were the characters.

Utter, utter fail, at least with my test image. Dismal.

WhatTheFont

WhatTheFont works in a very similar manner. Well, they were here first, so I'll let you figure out on your own who borrowed from whom.

One plus with WhatTheFont is that their OCR kicks in even before the character recognition step, so many of the characters are already filled in. You just have to go and fill in the missing ones and correct any mistakes that you find (I didn't find many).

As you can see in the screenshot above, WhatTheFont's algorithm clearly worked better for my image. For one thing, it actually identified the characters.

At the end of the process, it gave me five thin sans-serifs. They all seemed rather similar to the font actually used, but none of them were spot-on. Still, it was a nice effort, and I was fairly satisfied with the result.

Bottom line: WhatTheFont is, hands down, the clear winner. When the IdentifyFont algorithm can actually recognize the image, we might be able to run another comparison. Until then, WhatTheFont has nothing to worry about.

Tags: font, fonts, identification, ocr, services