To mitigate waiting on "preparing the recognizer" and reduce the delay to use by amending the option "listen immediately" to "start listening (recording) immediately when invocted & process the voice when the recognizer is ready.
This should be much faster than waiting for the recognizer by starting to record before the recognizer is ready. (On my Google Pixel 3 Sayboard takes about 2 seconds to "prepare the recognizer" every time I invoke it, but Googles "Recorder" and "Voice input" are ready very fast.)
This could be done in the back round and would probably be fairly seamless as the model would catch up to real time quickly.
To mitigate waiting on "preparing the recognizer" and reduce the delay to use by amending the option "listen immediately" to "start listening (recording) immediately when invocted & process the voice when the recognizer is ready. This should be much faster than waiting for the recognizer by starting to record before the recognizer is ready. (On my Google Pixel 3 Sayboard takes about 2 seconds to "prepare the recognizer" every time I invoke it, but Googles "Recorder" and "Voice input" are ready very fast.)
This could be done in the back round and would probably be fairly seamless as the model would catch up to real time quickly.
Would help mitigate https://github.com/ElishaAz/Sayboard/issues/30