First, make sure you have reviewed our general speech recognition tips.

Second, understand that a delay of a few seconds is normal. The speech recognition system needs to make sure that you are truly done speaking, and have not paused briefly, say to dictate a new sentence.

Longer delays are typically caused by background noise (which the app may think is speech) or a slow or unreliable internet connection (which leads to delays in converting the speech into text).  If a long delay occurs, wait for up to one minute before restarting the app.

There is also an advanced setting you can adjust that may fix the problem in certain cases.  Visit the "Speech Recognition (Advanced)" section of the settings, and try raising the voice activity detection threshold.