> ... I mean, this is what LLMs are. I suspect it's not an LLM for dictation? It's probably a network that converts sound to some phonetic tokens, then some k

> ... I mean, this is what LLMs are. I suspect it's not an LLM for dictation? It's probably a network that converts sound to some phonetic tokens, then some kind of probabilistic matching like a Markov chain (which would be computationally cheap locally)?