API for voice to text conversion.

Question

I have looked over the internet for voice to text conversion APIs. CMUSphinx, Android inbuilt API (Level 3), and pygooglevoice exist.

Which among them is the most accurate? (I am looking for an API to be used on a mobile platform).

Thanks.

Answer 1

Accuracy depends on the type of the speech you want to recognize.

• If you consider command and control type of the application with limited vocabulary, you can use a small acoustic model provided with CMUSphinx. It will be very efficient and accurate. It will not use internet connection as well which might be an advantages
• If you want to recognize open vocabulary text, mobile platform can't give you enough resources for such a complex task. For that you might want to use Google service through pygooglevoice

Answer 2

Old question but I stumbled upon it while searching myself. I ended upp using Androids Recognition Listener and it works great!

If you need an up to date go-through of it that also uses android architecture components, then take a look at this blogpost.