facebook/seamless_m4t · SeamlessM4T seems unavailable while the audio is paused.

That's my first time trying this translation model :-). I selected an audio source as the microphone and spoke some Chinese to it, with merely 2 seconds pause on it (13 seconds human voice, 2 second empty voice and 3 seconds human voice). However, the translated result is just about the first part, 13 seconds voice. The last 3 seconds, human voice also became an empty voice (it became noise actually). I don't know why it is happening, and how can I deal with this?

Attached files is translated English.