Open Grid

HTML5 Speech recognition in Chrome

Today I was testing new feature available in chrome which is speech recognition using proposed "speech" attribute (x-webkit-speech in Chrome). I was wondering if it is practical to speak into a input box as a way to write a blog post. Why you would want to write this way? Well, a lot of writers record audio of the things they want to convey and the use transcription services. But it cost money to do that and best tools and services are available in English only. Google Speech recognition works quite well in most languages.

Fiddle with speech attribute

To write this post I used this fiddle which added all the things I said into textarea field. (Try it out)

If I want to say something I have to click a microphone button inside the input box to activate the speech recognition (startSpeechInput() is not yet implemented in any browser). Then, I have speak clearly without significant pauses. Next, wait a little while so Google servers can recognize my speech. This time can be used to think about next sentence we want to say. To achieve best results it is good to say one sentence at a time.

When it's best to use feature

In the beginning it can be really painful and difficult to use this new way to enter text. Newcomers will probably be quickly disappointed. And if you write a lot of words per minute forget about it. Maybe for people who are unable to use keyboard in a efficient way it would be much better. When you use a keyboard your text is mostly grammatically correct and understandable. You don't have to correct your text if you are good writer so much. When I correct effects of speech recognition, I have to decipher most of the sentences.

One exception could be Smartphone and tablet interface. This kind of environment makes everyone a slow writer, so speech recognition comes in handy.

Conclusion

Speech recognition feature is not yet ready to be the best way to enter text into our text boxes. It is still frustrating experience.