I tried using google TTS to play on a MPD player, and found that it was cutting off a few seconds at the beginning and the end of the speech, although the MP3 file that was created was complete with no cut off.
When the TTS is directed at a device that generates the speech locally, such as Microft, it works fine without cutting off.
After doing some searching here and on the web, it looks like this is a problem that has been around for a while, and it happens with TTS from google, amazon poly, ibm, etc., and when it’s played back with any player that plays the stream, such as MPD, Logitech Squeezebox, Kodi, etc.
Since the MP3 generated is correct, and the players can play streams from the web without cutting off, it seems that it’s an issue with HA.
There are several suggestions here that basically add delays, or play a “silence” file, before and after the MP3, but these don’t work properly, and they are just a work-around.
This issue seems to have always been there (just like the issue with the contact bounce on the R-Pi GPIO pins) and never been solved.
I would like to see this corrected, so we can fully take advantage of the TTS functionality in HA.