Only part of audio transcribed

[NasonZ]

I have an hour long meeting which I would like to transcribe. I've attempted to do so with:

import whisper_timestamped as whisper

audio = whisper.load_audio("/content/Meeting Recording.mp4")

model = whisper.load_model("medium", device="cpu")

result = whisper.transcribe(model, audio, language="en")

My issue is that the result only covers the first 2 minutes of the meeting, what settings do I need to adjust to transcribe the entire meeting?

[Jeronymous]

Maybe there is a big silence gap after the 2 first minutes.
Can you try with option "vad=True"

[NasonZ]

There's no silence gap, but setting vad=True did help. I now managed to get 38 minutes out of 64 minutes, better but still not the entire meeting. Where the cutoff happened this time, there doesn't seem to be a significant silent gap, probably less than a second before one speaker replied to another.

Any ideas why it transcribe keeps prematurely ending?

[Jeronymous]

No, I have no idea why you experience this.
I would need the audio and the full set of options to reproduce and investigate

[Jeronymous]

This issue is really strange. @NasonZ Is there a way you can share the audio?
(either by linking the zipped audio here, or sending it to hello_at_linto.ai)