When model config defines phone_silence mapping, each sentence is broken up into phrases at the phonemes that need added silence. The phrases are synthesized independently and silence audio samples are inserted between them.
Only tr_TR-dfki-medium defines phoneme_silence at this time
https://github.com/rhasspy/piper/blob/e5cb84cd5c3cb93eff9398264d5a7da5122ff21a/src/cpp/piper.cpp#L508
When model config defines phone_silence mapping, each sentence is broken up into phrases at the phonemes that need added silence. The phrases are synthesized independently and silence audio samples are inserted between them.
Only tr_TR-dfki-medium defines phoneme_silence at this time