Re: How does vad work?

Oops, that there *may* be consonant sounds prior to the detection of audio before the trigger point.

Also, I turned the –t option all the way down to 1 and I’m *still* losing the first half of “guess” in “guess what”.  That’s just sad.

Damon

I’m already using norm along with vad.

What’s frustrating is that vad seems to perform *less* effectively at stripping silence than the “silence” option does.  The p option for vad seems to be just a shot in the dark that there *may* be vowel sounds prior to the detection of audio before the trigger point, and trying to capture “gue” out of “guess what” with –p seems rather silly, given how long of a duration that missing fragment is.  (Yes, that’s how much is being stripped by vad!)

My goal is to tightly constrain the audio clip to the detected audio, and remove the silence from both ends.  Having –s and  –p options for “silence” would help enormously, since the only time “silence” seems to lose anything is with a few edge cases like certain initial consonants (c, h, k), certain labial stops at the end (-t, -th, -d), and words ending in a nasal consonant (m or n) that’s relatively quieter than the rest of the audio.

Damon