HTML5 Canvas powered, with a flash fall-back so it even supports ie6 (very slow on it though). Supports Twitter, Facebook, and pasting in anything you like.

My main use of "word cloud" generators (wordle.net in particular) is to take large texts, show all words except the most common, and - this is key - highlight particular words.

Prime example: http://www.wordle.net/show/wrdl/1507537/Reservoir_Dogs (the easily offended may refrain)

To do this with Wordle, I have to use a word processor & spreadsheet to count the words, and specify the RGB color for each (all the same except the few I want to stand out, however common/rare). The "common words" removal becomes problematic when the words to highlight are in that removal list (say, when rendering a speech by the Obama to highlight frequency of "I", "my", "me", etc.). Ergo, would be nice to render all words no matter how rare, specify default and per-word colors without marking all, and indicate specific exemptions from the "common words" list.

Really cool. And you have also shared the technology used in the site. May be you could blog about the development life cycle detailing some of the rationale for selecting a given tech. Kind of like making of Wispy. I am sure many of the us will get benefited out of that.

I'm planning on doing that, and open sourcing quite a bit of the tools/libraries I had to build myself. The contact page is a little fancy in that it sniffs your browsers feature set via javascript and sends it on to me when you send a message, a huge help with tech support :) Its up on github and linked to on the faq page.

I have a few feature requests; mainly because I noticed missing 3 letter words in my own cloud:

  * Add handling for hashtags (overriding the min character length)
  * Slightly more complicated: consider adding click detection on mentions, replies, and hashtags to link to their perspective urls. This would be a very neat addition.

http://www.quora.com/HTML5/Is-it-possible-to-add-links-to-HT...

At present wispy ignores all 3 letter words, apart from a small list considered "non boring", it also ignores plenty of longer words that are considered generally meaningless, The hashtag idea is good, ill look into that, consider it on my todo list :)

Click detection is something I have been thinking about, its not as hard as you think as the general coordinates of each word are stored when building the cloud in the first place.

My twitter account (@vraa) is not protected. But when I tried, I got an error message "Whoops, either that twitter user does not exist, or their account is protected, try someone else.". tried both vraa and @vraa without success!

may be a bug or twitter quota has exceeded?

It uses jsonp to talk to the twitter api, so if its a rate limit, its on your own ip, are you in a large company or uni by any chance? Otherwise it might just be a slow connection, you dont really know when a jsonp call has failed, so you just have to time it out after a while and assume thats what happened, happens to me every now and again when using it from a phone. If its none of the above, please send me a message using the sites contact form, it tells me a lot about your browser in the process and I can try and replicate the problem.

Try changing the case of the letters, I had the same issue.

Nifty. It always bothered me that Wordle required Java.

Thanks, that's pretty much why I built it, its still not as nice as wordle though, but I had to make a lot of feature/performance trade-offs to get it running at a decent speed.

nicely done. always enjoyed playing with Wordle.