That's partially due to Google's approach. It appears that they're building a single model for every acoustic model, language, and domain. While that works as a

That's partially due to Google's approach. It appears that they're building a single model for every acoustic model, language, and domain. While that works as a general purpose tool, it's less useful as you go down the rabbit hole. Side question.. how do you host your podcast? I'd love to chat if you get the chance. My email is in my profile.