So after playing around with the page, I just started saying random things to see how the speech API would fare with my atrocious accent. That was interesting, until I said "the quick fox", and it interpreted ... bleeped speech?

    2015-06-20 03:02:49.109 say_restyle.js:7 Speech recognized: the quick f****
    2015-06-20 03:02:49.110 say_restyle.js:7 Speech recognized: the quick Fox
    2015-06-20 03:02:49.110 say_restyle.js:7 Speech recognized: quick f****

What the fox?

I noticed that while I was developing the script. I'm using annyang -- https://github.com/TalAter/annyang -- to interact with the web speech API, and it doesn't look like annyang is the one censoring the results, so it must be ... the API itself?

That is unbelievably messed up. The fear of swear words in american culture actually messing with APIs. I really hope this isn't what it looks like and someone is playing a prank somewhere.

Yes - in particular, it's done server-side by Google. It seems that there is a config option to turn it off. http://src.chromium.org/viewvc/chrome/trunk/src/content/brow...

Am I really the only one freaked out by this?

No. The idea that your voice is being recorded and sent to Google is something that a lot of people might not want.

I think you missed the point.

Can you say specifically where for the stupid people like me? I looked through and did not see anything.

Interesting that they're using FLAC as the encoder... Seems like it would be very memory intensive.

I envision this to be a great little browser extension for those pesky team / site presentations.

Can you make that text yellow? Sure, tell it to do so and see how stupid it looks for yourself!

I have no idea what is going on, am I suppose to just speak to the page? Because it does nothing (safari)

EDIT: apparently visiting https://www.google.com/intl/en/chrome/demos/speech.html it tells me safari is not supported :(

Why are these cool things never in Firefox? As I don't use Chrome I often run into demos that don't work.

The Web Speech API in Chrome is implemented by calling Google Speech Recognition APIs. Firefox would have to pay for an API key to use the same service.

There has been some work for client-side speech recognition in browsers though, for example: http://syl22-00.github.io/pocketsphinx.js

Yet the project itself indicates it works in safari... :/

Now that the speech API is accessible to almost all developers, I really see this being a very helpful thing for developers while they work on web sites.

Want to try multiple styles on a mobile app (phonegap/cordova) without recompiling it? Need to make "fuzzy" css changes to gradient colors ("make it darker") while experimenting?

Then, in a little while we will be able to roll this same powerful API out to our users.

seems more tedious than open up dev tools. too much variation to be of general use, probably depends on certain markup like bootstrap. innovative idea though.

This is quite interesting and is very close to what we are doing at http://www.dhi.io. Though the difference is that , there is actually an AI sitting in between the user and the editing layer - interpreting the users commands, making the edits etc.

I've seen some neat projects using the speech API, including this one, but the recognition seems way too bad. Where does the recognition engine come from? Do the browser vendors make it?

Really cool! And it actually works!

Seems like 'select its first child' doesn't select the first child, but the second.

This is really cool!