Half a croissant, on a plate, with a sign in front of it saying '50c'
h a l f b a k e r y
Not just a think tank. An entire army of think.

idea: add, search, annotate, link, view, overview, recent, by name, random

meta: news, help, about, links, report a problem

account: browse anonymously, or get an account and write.

user:
pass:
register,


                                               

Shazam For Fonts

Point your phartsmone at a font and it tells you what it is
  (+9)(+9)
(+9)
  [vote for,
against]

Shazam uses alien technology (okay, quite clever technology that is a centre of attention right now on the blog circuit) that can decipher which choon you’re listening to.

Fonts are a similar thing, where people see one and wonder what it is. There’s a flickr group for typeface ID-ing, and it’s entirely driven by human recognition and participation. This would put a stop to such futility.

Ian Tindale, Sep 26 2010

http://kottke.org/10/09/how-shazam-works [Ian Tindale, Sep 26 2010]

What The Font http://new.myfonts.com/WhatTheFont/
Google is your friend {Other search engines also} [Dub, Sep 26 2010]

Identifont.com http://www.identifont.com/
You help describe it [Dub, Sep 26 2010]

FontMatch http://www.stretche...match/fontmatch.php
[Dub, Sep 26 2010]

Cheese or Font http://cheeseorfont.mogrify.org/play
Can you tell the difference? [Dub, Sep 26 2010]

[link]






       Could this be done without buying a license for the fonts it recognizes? (Question, not criticism. I would buy this, if it existed and I could afford it.)
mouseposture, Sep 26 2010
  

       How would it achieve this? There are quite subtle differences between many type faces ie Helvetica/Univers.
xenzag, Sep 26 2010
  

       //Could this be done without buying a license for the fonts it recognizes?//   

       Why not? I suspect that the database would need to hold a certain number of parameters derived from each font, but not the font itself. If I measure a "w" and a "v" in Old French Bastard Italic, and find that the former is 1.8 times as wide as the latter, surely that information is not copyright?   

       //How would it achieve this?// I suspect it might be very simple, using any of a wealth of possible parameters. For example, divide the text up into individual characters (no need to recognise them), and weed out the ones that are identical. So you now have a number of different characters. Then measure the width of each character. Then calculate the ratio of each width to each other width. So, if you had a sample containing 20 different characters, you would get 400 ratios.   

       Then just find the font which has 400 ratios that best match your "fingerprint".   

       Or whatever.
MaxwellBuchanan, Sep 26 2010
  

       I have a feeling that it's going to be more difficult than that. There are numerous weights and other very subtle variations within each typeface never mind between whole fonts. What does the camera photograph and how does it translate, for example, a font printed over a half-tone image?   

       Interesting problem for a phd student.
xenzag, Sep 26 2010
  

       Well, Shazam probably has issues identifying a tune played over a brass band, but point taken.   

       However, I still think that some sort of simple internal ratiometric fingerprint will distinguish most fonts.
MaxwellBuchanan, Sep 26 2010
  

       And what would the price point be? [Ian Tindale] seems to have in mind something cheap, for the masses, but it could be sold dear to the owners of fonts. They would then launch a fleet of web-crawlers harvesting unlicensed instances. A tidal wave of "Cease & Desist" letters would follow.
mouseposture, Sep 26 2010
  

       Perusque mandate ille utrescem?   

       I don't see why identifying a font would be a problem in and of itself. Presumably Shazam doesn't get sued simply for telling you the name of a tune you've heard?
MaxwellBuchanan, Sep 26 2010
  

       <Bickle> "You talkin' to me?" </Bickle>   

       No, but Shazam is no doubt useful for detecting unlicensed copies of that tune. The point being that this idea would be used to identify targets for such suits, not that it would itself be a target.
mouseposture, Sep 26 2010
  

       I don't understand how Shazam helps to detect unlicenced copies (eg, Shazam tells you it's "Whale Sundays" by the Trilobites; then what?), nor how a font identifier would help you detect unlicenced copies (eg, it tells you that the font is "Angina Sans Serif"; then what?).
MaxwellBuchanan, Sep 26 2010
  

       I dunno, I may be missing something, but here's what I thought: It doesn't detect unlicensed copies, but, incorporated into a web crawler, it detects copies, automatically, and on a large scale. Which is a help in detecting copyright violations.   

       //Shazam tells you it's "Whale Sundays" by the Trilobites; then what// Then you check the web page your crawler found it on, and it's a student homepage in the University of Southern North Dakota at Hoople domain, along with other copyright material and a helpful hyperlink to Pirate Bay. Then you send the cease & desist letter.
mouseposture, Sep 26 2010
  

       Essential when duplicating legal documents for illegal purposes. [+]
daseva, Sep 26 2010
  

       //Then you check the web page your crawler found it on//   

       Ah, how do you do that with Shazam? And likewise, how would you do it with this font identifier?   

       One of us has missed the point here - the font identifier needn't trawl the web to find a matching font; it just has a library of font parameters, so it can tell you what font you're looking at.   

       Or maybe I've musinerstood the idea here.
MaxwellBuchanan, Sep 26 2010
  

       How does it "know" that it's looking at a font when you point it towards a billboard with a mixture of text and images?
xenzag, Sep 26 2010
  

       A web crawler simply traverses links, visiting webpages. That technology is well-baked and available off-the-shelf. What it does with each depends on what filter you associate with it, and, if your needs are specialized, you might have to write your own filter. A simple filter might scrape the page for email addresses: that's easy. Deciding whether any of the audio files linked from the page contain a song whose copyright you own ... that's much harder, but Shazam has solved the problem for you. The result is a shortlist of webpages, for inspection by a human. You don't check those out using Shazam, you check them using an ordinary browser.   

       In the case of fonts, this would only get you as far as a photograph (or scan) of a copyright violation, and further work would be required to track it down in meatspace.   

       My point is, there's an easy and a hard part to trawling for copyright violations in this way; the easy part was solved long ago, and this idea solves the hard part.   

       And what are you on about re: Angina. That's a serif font. Maybe you mean Infarct?
mouseposture, Sep 26 2010
  

       300,000+ fonts, 130 characters in a full font set, 18 differentiation points per character... this may be a lot bigger project than anyone realises.   

       However, this would be absolutely invaluable to designers and printers. Bring it on!
infidel, Sep 26 2010
  

       //300,000+ fonts, 130 characters in a full font set, 18 differentiation points per character...//   

       Yes but, as noted, you don't need all that data to find a match. There are 6 billion people in the world, each different from the other by a vast number of points of detail. Yet a hundred or so bits of information are enough to identify any individual. Likewise, there are probably more than 300,000+ tunes, each with hundreds of notes played on dozens of different instruments, yet Shazam works OK.
MaxwellBuchanan, Sep 26 2010
  

       I still want to know how the fonts get separated from their backgrounds. Scanning an individual letter on a white background is very different to photographing something printed on a menu stuck to wall on the other side of a room.   

       I also think there are more than 300,000 fonts, then are weight variations, ie bold, light, italic, underlined etc, then there are colours..... red against magenta etc.
xenzag, Sep 27 2010
  

       Yes, but who says this has to work under all conditions? Shazam, I imagine, doesn't work if there's a huge amount of background noise. Likewise, this tool might be excused if it failed under poor conditions. If it managed to recognize fonts photographed half-decently, in black on a light background, it would be doing pretty well.
MaxwellBuchanan, Sep 27 2010
  
      
[annotate]
  


 

back: main index

business  computer  culture  fashion  food  halfbakery  home  other  product  public  science  sport  vehicle