In web or desktop application, what is an acceptable delay for speech recognition processing?

That is, how long will it be before users get annoyed or dissatisfied with having to wait for speech recognition?

Links to relevant articles or research (if it exists) would be appreciated.

  • 2
    Hi and welcome to User Experience! What does your application do? Does speech recognition power real-time interaction (voice commands) or is it simply dictation? They will probably have different thresholds. If I'm dictating a letter and I'm blurting out a stream of consciousness, I'm less fussed if it's slow than if my Google Home lags and takes 20 seconds to turn my lights on. – msanford Feb 8 '18 at 14:45
  • 1
    @user3724327 Appreciate the edit but Mike has not yet clarified that that is his intent. – msanford Feb 8 '18 at 15:33
  • 1
    Agreed, and rolled back to original. @user3724327 Until Mike clarifies, it's not accurate to add/remove information about the question. – Alan Feb 8 '18 at 15:46

You can refer to this research and its' findings: An Experimental Evaluation of Apple Siri and Google Speech Recognition. Page 4 in particular.

The reason I am suggesting this research is the competition of two global corporation that is based on great research in UX.

In my opinion, 5 seconds is an appropriate delay an average modern application user is patient enough to wait.

  • Great linking to the research paper! So +1 from me, but, I'd add that it's from 2015, which is ancient history as far as speech recognition is concerned. This is only an opinion, but if I have to wait more than 500ms for speech recognition to begin, I assume something had gone wrong. My keyboard recognizes speech nearly in real-time, as does my Google Home (the action delay is related to control API calls to the respective devices and time to return search results, not the speech recognition itself). Google Home / Echo / Siri / Cortana provide keyword feedback within 200 ms. – msanford Feb 8 '18 at 15:00
  • My Alexa takes about 3 seconds to answer. I noticed that it takes longer when my router is loaded with a lot of other devices (3 phones, 1 media server, 2 laptops, apple tv, 1 smart tv and 1 desktop). I once used them all at the same time, so my line load was ~80%. Media server was undergoing some serious overloads. I asked Alexa what was the weather like and it took ~10 seconds to respond. Thinking of this, I believe that those numbers in 2015 are related to internet connection speed three years ago. – Ivan Venediktov Feb 8 '18 at 15:16
  • Note that "answer" relates to the second part of my comment: (the action delay is related to control API calls to the respective devices and time to return search results, not the speech recognition itself). Yes, that part can be long, but what's the delay from "Alexa" / "Hey Google" / "Hey Cortana" to it waking? Much shorter. That also involves speech recognition (albeit onboard). Getting the actual full response from a complicated query involves many more moving parts than just the speech recognition. – msanford Feb 8 '18 at 15:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.