Force silence using built-in VAD during encoding

[vadimkantorov]

Hi!

Is there an option for opusenc to reset to exact zero the frames without detected speech during encoding (leveraging this built-in VAD)? Is it possible to control the used thresholds from the opusenc frontend?

Do encoded frames contain a silence bit? (if so, these frames can be skipped during decoding as they would not contribute to speech recognized text further down the pipeline) I've read RFC and it seems so! This is also relevant for stereo and multi-channel call recording files as during decoding most of frames are silent when only one person is speaking.

Thank you :)

[vadimkantorov]

New --channels discrete can be used to force uncoupled encoding, but it would still be nice to have a mode for super-storage-space-efficient silent frame encoding (using VAD or by passing some threshold) of long effective silence (e.g. one silent person listening to another person speaking)

[vadimkantorov]

Related on DTX:

• Option for opusenc to omit DTX/silence frames from output completely and provide options for controlling aggressivity of silence detection #89