Gaps in audio playback when using MaryTTS

Having a problem when using MaryTTS… when I give her a phrase, there are gaps in the output, almost like a buffer is getting overrun. All the output is spoken, eventually, but it’ll be broken up… in the mid…dle kind of… like… this.

But when I hit the maryTTS via a web browser, save the output to an mp3 file, throw it in the (openhab)/sounds directory, and use playSound(file.mp3), it plays perfectly with no gaps in the output whatsoever!

So I’m looking for ideas on what could be going wrong here. Mary is running on the same box (an Atom box), and the cpu load is about 0.5 when Mary is generating output. So I should still have plenty of horsepower left over. (It’s about 0.2 to 0.3 when playing the mp3 file.)

Doesn’t look like memory is an issue, either. (Taken while mary was processing something to say.)

~$ free -h
             total       used       free     shared    buffers     cached
Mem:          2.9G       2.0G       942M       9.0M        82M       804M
-/+ buffers/cache:       1.2G       1.8G
Swap:         1.0G         0B       1.0G

Looking for ideas or things to check why mary would stutter. Thanks! (openhab 1.8.1, maryTTS 5.1.2)