Timestamp skew in long audio

[Tortoise0Knight]

I'm trying to transcribe an audio from a movie, which is about 2.5h. I'm using large-v3

I've noticed timestamp skews in some parts of the scripts, which are usually accompanied by a series of duplicated words.

As far as I know, chunking is automatically done by whisper with long audio. Still is there any connection between the problem and the length of my audio?

[glangford]

• Problem of inaccurate subtitles for long videos #1811