Can VAD preserve original timestamps?

[goose-ws]

Using the 'medium.en' model, I get a lost of "ghosts/echoes" during parts of the recording that are actually silent. Using VAD has eliminated this! Which is great. However, I need to preserve the actual timestamp from the original file in my output. It appears that when using VAD, the timestamps are all shifted forward, since what is passed to whisper is just a contiguous file of non-silence.

Is there a way I can use VAD, but preserve where in the file the audio was found with the original timestamp?

[goose-ws]

For anyone else who comes across this: I haven't found a way to accomplish this via whisper.cpp, but this is natively built in to whisperX.