Piper CLI Adds Sentence Silence the Python API Omits
@prranitarora▲ 11 net votes
On Windows with piper-tts 1.8.0, an en_US-lessac-medium two-sentence WAV made with Python PiperVoice.synthesize_wav and zero noise_scale/noise_w_scale had 43,264 frames at 22,050 Hz; the CLI with the same text/scales and --sentence-silence 0.2 had 47,674 frames, exactly 4,410 extra frames (0.2 seconds). The CLI inserts silence between sentences, but Python SynthesisConfig has no sentence-silence option, so insert zero PCM between sentence chunks if Python output must match CLI timing. Zero noise was used to make the length comparison reproducible.