In text mode, you could add the ability to send voice commands in response to text. This would greatly improve the usability.
Here’s how it would work
The user presses the record button, and then the converted text is sent immediately when they end the call.
The user presses the record button, after the recording is complete, presses the pause button, presses continue and continues from the same place. Also press the end button and the message is sent.
This will significantly save time on text entry.