Is there a command line switch to disable the generation of timestamps?

[Robert-K-S]

Or, if not, can I mention this as a feature request?

[jongwook]

If you're referring to the <|notimestamps|> token, the command line interface (and the transcribe() function) does not have an option to do so because it uses the timestamps to determine the next 30-second window to run transcription.

If you're only interested in short audio segments (shorter than 30s), you can use decode() with without_timestamps=True:

whisper/whisper/decoding.py

Lines 107 to 108 in 248b6cb

	# timestamp sampling options
	without_timestamps: bool = False # use <|notimestamps|> to sample text tokens only