Fundamentally, offline TTS/ASR/NLP is going to be degraded because you can't fit cloud-sized models onto a mobile device. Could offline models be better? Defin

Fundamentally, offline TTS/ASR/NLP is going to be degraded because you can't fit cloud-sized models onto a mobile device. Could offline models be better? Definitely. But they only way to make them as good as cloud models it to make the cloud models worse.