Silent audio is recognized and the timestamp is shifted thereafter.

[Shin-ichi-Takayama]

Hello everyone.
I entered voice into Whisper and checked the recognition results.
Then it recognized the voice in silent time. Also, the timestamps of the subsequent recognition results were off.
Is there any way to address this issue?

I am attaching the audio and recognition results for your reference. The language is Japanese. Model size is medium.
test.zip