I found Tasker speech to be pretty clunky, inaccurate and minimally supported. If it's for personal use I have google recorder on a Pixel 7 with GrapheneOS and it's STT is outstanding. It'll download the offline engine and then you can kill it's network and rip out google serices and it works fine (for now). Might be something to explore... If you find something good please reply, I'm thinking alon the same lines, but was just going to send the audio directly and do stt on the server. I know home assistant is planning voice command next year, might be something helpful there. You could also try using the Termux environment to spin up whatever linux engine you like.