Splits long 16 kHz audio at its quietest 50 ms between minSeconds and maxSeconds, so cuts land in pauses
rather than mid-word, then trims each window's silent edges and drops windows with no sound. Moonshine tiny
returns nothing for a window that opens with seconds of silence, and trimmed windows also give timestamps
that start at the speech. It is plain RMS with no model, so an hour of audio splits in milliseconds.
Splits long 16 kHz audio at its quietest 50 ms between
minSecondsandmaxSeconds, so cuts land in pauses rather than mid-word, then trims each window's silent edges and drops windows with no sound. Moonshine tiny returns nothing for a window that opens with seconds of silence, and trimmed windows also give timestamps that start at the speech. It is plain RMS with no model, so an hour of audio splits in milliseconds.