Similar visual search for LaTeX characters: http://detexify.kirelabs.org/classify.html

First thing I thought of when I saw Shapecatcher was Detexify. The similarities are obvious, it was most likely his true inspiration, or at least part of it, yet he didn't mention in the about page.

I just ran across detexify today while trying to remember what \gtrsim was called... Quite a coincidence to find the unicode equivalent on HN a few hours later.

I thought it said "unicorn". I was dismayed to find that my beautiful, refrigerator-worthy rendering of a unicorn was interpreted as a few Arabic characters, some arrows, and a domino tile.

Yet it knows USSR: ☭ recognized just fine.

Great concept, but 1 for 5 in recognition. Matched dollar sign. No euro, no pound, no ampersand, no pi? Maybe I'm a horrible artist, but 49 suggestions (some that look like pi) and no pi?

One thing they could do to help alleviate this is favor the Unicode section(s) associated with the visitor's language, as well as punctuation. That way we'd be less likely to get results like "Ogham letter nion: ᚅ" for drawing pi.

Pound (£) sign was fine for me, even though I actually drew a lira sign (₤ - is has two crossbars). Hash (#) was ridiculous, though:

Tifinagh letter tuareg yazh: ⵌ

Vai syllable pu: ꖛ

Equal and parallel to: ⋕

Viewdata square: ⌗

I tried the ae ligature, long s, sharp s, Russian zha, Greek capital gamma, and a heart, and of those only the gamma was in the top five (it was #2); zha, ae, and heart weren't even in the list, and long-s and sharp-s were waaay down the list. (Actually, I just tried it with a better-drawn sharp-s and it didn't even turn up.) In some cases the top choices really do resemble the letter I was looking for, so that's fine, but there's a lot that shows up that resembles the canonical form a lot less than my drawing does.

So, not quite ready for prime time, I think. Cool idea, though.

Another datapoint: I had no trouble with any of those.

Pi took me a couple of goes, and when I drew a somewhat narrow ampersand, it gave me a few erroneous results.

Both were successful, and put the desired character at the top of the list. All the others worked first time.

It would be excellent if it could be made to learn, with users being able to say "that's what I meant"

Am I the only one who tried "ಠ" and couldn't get it to work no matter what?

I tried a snowman ☃ - it's pretty low on the list but you'll find it with a crude but detailed drawing.

The secret is adding the snowflakes.

Is it in a free font? http://www.alanwood.net/unicode/kannada.html suggests maybe not.

Yes, it's in Dejavu Sans, which is what Shapecatcher uses. I'm guessing the Kannada block just isn't in the database yet.

It isn't included in Dejavu Sans. I (and others) had to install a different font to be able to see it.

You are right. On my system it's being automatically substituted from another font, even in character map applications.

I really wish font handling and rendering were a bit simpler to understand and troubleshoot in Linux.

Recently I had an annoying problem where some web pages weren't rendering the fonts correctly. It turned out the reason was a font config file for a Chinese bitmap font was causing firefox's font rendering to alter - both chromium and another browser were working as expected (according to the css rules). Tracking down the problem took several weeks and several dead ends and I only stumbled on the answer by accident anyway.

Very cool. Would be useful to see a section of neighboring glyphs in the area of a selected glyph (I want to find the spade suit and see some other suit glyphs too)

I was recently searching for a unicode Thumbs Up, and came up empty handed (there is none). So I drew one in this thing, and got some useful approximations at least:

ຢ ථ

I also came across this one, describe as "weary cat face":

🙀

Seriously, Unicode? You've got "weary cat face", but no thumbs up? I am disappoint.

I noticed that the second attempt had more accuracy than the first since I was able to see a similar character (white rook) and then modify my crude drawing to eliminate extra information that was adding white noise to the search. Nice work!

I tried really to get the thumbs up character, but no luck: http://i.imgur.com/YoWFh.jpg

Handy tool anyway :)

Apparently it can't do chinese or Japanese characters.

It says on the page

>Note: Japanese, Korean and Chinese characters are currentenly not supported, but I'm working on supporting them!

Jim Breen's page can do this, though looking up by radical is faster if you know how to do it.

http://www.nciku.com has a really good character recognition engine. I've never had it not find exactly what I needed.

Agreed. I randomly tried イ, ア, 月, 四, and 行. It pulled back tons of results for each of them, but not a single list contained the correct result.

Yeah, I encountered the same issue. The only thing I can think of is that I'm drawing them in the handwritten style, but the recognizer probably needs them to be in the digital font style.

For kanji try this: http://kanji.sljfaq.org/

Or runic... but I guess I'm showing my age there.

This is top notch. Any chance you would make this so that you could embed it in another website - maybe for a cost/month? because I can think of some sites right now that could really use this.

Very useful, fun and even has research value. Draw a circle and find a lot of emoticons. I couldn't find any of the letters from my mother tongue though, hopefully they will be added.

It is astonishing how many types of vertical dashes there are.

This is cool! I'd love to look into using the recognising algo in a rewrite of motion captcha. Similar concept.

Great site.

Just used it tonight... encountered a letter psi 𝜓 in a recaptcha form and it helped me find it. ;)

Hundreds of people just checked to see if there was a Unicode character that resembles a penis.

This is a genius name. Sounds catchy, but still nails what the product actually does.

I'm drawing some seriously inappropriate things, and this site does not disappoint.

Bookmarked! Wow, quite an amazing product. How long did it take you to develop?

Great. Tried some Tamil characters. It was able to recognize most of them.

Couldn't get it to recognize the "look of disapproval" character :-(

Did not recognize my unicode snowman, but it got pretty close: Ṏ Ổ ồ

I actually got the snowman as the 8th result when trying (and failing) to draw the unicode character "pile of poo" U+1F44D: http://i.imgur.com/VBvK6.jpg

It does expect a fairly unorthodox snowman: http://i.imgur.com/IktTl.png

Great stuff, unfortunately it doesn't catch my Devnagri letters.

I got one right after 10-15 tries. http://i.imgur.com/4SJsf.png

But it does get almost all Malayalam letters I tried.

Awesome work.

Really? It got my Devanagari letters surprisingly well.

I tried drawing a lowercase 'a' twice, and it couldn't find it.

I tried drawing the symbol for pi 3 times, and in dozens of results did not see the character for pi. Odd.

I got pi on the first try: http://i.imgur.com/IPyn7.jpg

Very nice! It caught my terrible rendition of the interrobang.

I can't, for the life of me, draw and get it to recognize "ñ"

Strange. Worked first time for me: http://i.imgur.com/JD2qb.png

Touch support would be nice to have.

When that link was posted, the site was non-operational.

⚛ Jimmy Neutron! ⚛

This is amazing.

I tried some astronomical symbols, like ♃, but they don't seem to be supported yet.

One of the first things I tried was Mercury ☿ and it did work (even first match) - and that's despite my drawing skill.

Interesting. I actually tried a few different variations of Jupiter's symbol (like making the curve more or less exaggerated, higher and lower, etc.) and couldn't get it to work.