Error "Wrong tts file format" when calling tts.cloud_say to MPD media player

I have a MPD installation on a raspberry pi 3b (Raspbian Bullseye). I use it as a media player for HA. In general it works but sometimes I’m getting the following error when calling tts.cloud_say. Any ideas?

Logger: homeassistant.components.tts
Source: components/tts/
Integration: Text-to-speech (TTS) (documentation, issues)
First occurred: 09:58:37 (3 occurrences)
Last logged: 09:58:37

Error on load tts: Wrong tts file format!

I am having the same error. Using MPD on another computer in network.