They all use Sphinx for speech recognition under the hood, which sadly works very poor for me.