One of the frustrating experiences of using Google’s Gboard for voice typing is the occasional long pause before endpoint detection and populating the results on the screen. I’ve learned to be patient, meaning waiting up to five seconds for the text to appear after having stopped talking.
The reason this delay is more likely to lead to an aborted input (user pressing the microphone button to stop listening) is that there is no indication that Google is actually working on a result. If the outline of the text (similar to searches using Google Assistant) was shown on the screen, it could at least indicate that this wasn’t some other failure.
Compared to typing on a touch screen, voice input is significantly faster. However, the experience needs to be more consistent for it to be a viable replacement.