Internally, we're making some changes to the VAD API to avoid footguns - namely ones where you get the end of a segment and then the start of the next segment and have to maintain some internal state tracking, which gets more onerous the longer the chunks of audio you provide (providing a whole file would be a bit painful).
This sees how the change in the code looks in an actual engine and sees if it simplifies things. It doesn't attempt to access any of the buffer manually so doesn't fully test the API but we can see it appears to simplify things (before and after comparison is currently a TODO because it's late here)
Internally, we're making some changes to the VAD API to avoid footguns - namely ones where you get the end of a segment and then the start of the next segment and have to maintain some internal state tracking, which gets more onerous the longer the chunks of audio you provide (providing a whole file would be a bit painful).
This sees how the change in the code looks in an actual engine and sees if it simplifies things. It doesn't attempt to access any of the buffer manually so doesn't fully test the API but we can see it appears to simplify things (before and after comparison is currently a TODO because it's late here)