4 ms·

Show HN: Edit this webpage with your voice

73 points by jawns 11y ago

Xeoncross 11y ago
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.
- x5n1 11y ago
  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.
ConAntonakos 11y ago
This is really cool!
scrollaway 11y ago
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?
- jawns 11y ago
  I noticed that while I was developing the script. I'm using annyang -- https://github.com/TalAter/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?
  - scrollaway 11y ago
    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.
  - TD-Linux 11y ago
    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/browser/speech/google_one_shot_remote_engine.cc http://src.chromium.org/viewvc/chrome/trunk/src/content/brow...
    - scrollaway 11y ago
      Am I really the only one freaked out by this?
      - userbinator 11y ago
        No. The idea that your voice is being recorded and sent to Google is something that a lot of people might not want.
        scrollaway 11y ago
        I think you missed the point.
    - chejazi 11y ago
      Interesting that they're using FLAC as the encoder... Seems like it would be very memory intensive.
    - aw3c2 11y ago
      Can you say specifically where for the stupid people like me? I looked through and did not see anything.
sr_banksy 11y ago
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!
baby 11y ago
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 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.
- TD-Linux 11y ago
  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 http://syl22-00.github.io/pocketsphinx.js
- mgkimsal 11y ago
  Yet the project itself indicates it works in safari... :/
borplk 11y ago
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?
ganarajpr 11y ago
This is quite interesting and is very close to what we are doing at http://www.dhi.io 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.
rolfvandekrol 11y ago
Really cool! And it actually works! Seems like 'select its first child' doesn't select the first child, but the second.

73 points by jawns 11y ago

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.

x5n1 11y ago
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.

ConAntonakos 11y ago

This is really cool!

scrollaway 11y ago

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?

jawns 11y ago
I noticed that while I was developing the script. I'm using annyang -- https://github.com/TalAter/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?
- scrollaway 11y ago
  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.
- TD-Linux 11y ago
  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/browser/speech/google_one_shot_remote_engine.cc http://src.chromium.org/viewvc/chrome/trunk/src/content/brow...
  - scrollaway 11y ago
    Am I really the only one freaked out by this?
    - userbinator 11y ago
      No. The idea that your voice is being recorded and sent to Google is something that a lot of people might not want.
      - scrollaway 11y ago
        I think you missed the point.
  - chejazi 11y ago
    Interesting that they're using FLAC as the encoder... Seems like it would be very memory intensive.
  - aw3c2 11y ago
    Can you say specifically where for the stupid people like me? I looked through and did not see anything.

sr_banksy 11y ago

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!

baby 11y ago

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 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.

TD-Linux 11y ago
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 http://syl22-00.github.io/pocketsphinx.js
mgkimsal 11y ago
Yet the project itself indicates it works in safari... :/

borplk 11y ago

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?

ganarajpr 11y ago

This is quite interesting and is very close to what we are doing at http://www.dhi.io 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.

rolfvandekrol 11y ago

Really cool! And it actually works! Seems like 'select its first child' doesn't select the first child, but the second.