DS Alignment boundary issues.!

I want to share my experience with you.
As there is the only possibility of 10 ms , 20 ms, 30 ms limitation of fragment length.
No matter, I choose 30 ms or 20 ms or 10 ms.
Or, if I try with some alteration of webrtcvad library to reduce it further like 1 ms / 2 ms / 4 ms etc or so.

There is boundary issue. Always. The data while transcribing is always missing. Is there any bottleneck of the algorithm? Or maybe there is any altered way to handle it better. Because If we are handling it with the time based, it will always be like this. If there is any vocal based way, maybe could handle better.

Still I am also trying to find way. But you are top experts. You can better comment and suggest the way.