Prompt makes more segments silence

[conan1024hao]

After applying prompt (about ten words, looks like word1, word2, ..., word10, to prevent misspellings), Whisper did recognize some words more correctly, however, the results' overall WER became about twice.

I found that some segments, which Whisper can recognize correctly before applying the prompt, are completely silence now, which made WER so high.
Adjusting beam_size and no_speech_thershold can not solve this issue completely.
Both large and small model show this issue.

Possibly relevant discussion: #1594

[conan1024hao]

The silence issue was solved after merging these separated utterances into one utterance.