I actually intended the two to be displayed even when --verbose is False, thinking it'd help the user notice any language detection errors and show the progress/ETA without flooding the console.

@jongwook well, ETA is disabled when verbose=True, which is misleading as it seem to me. Also, the language tag is returned as a part of the result. Do you really need to print it if you return it?

The language tag is returned in the result, but for those calling whisper in the command line (as it seems to be a popular use case) the detail is lost.

Regarding ETA: that's true; it's not the most intuitively named options but I thought both formats have merits (like below) and thought it'd be nice to have these verbose/concise output options, but probably better not mix up those two.

We can maybe make it an integer level for verbosity if a totally silent output is desired by many ..

@jongwook
How about str verbosity like "full", "minimal" and "silent"? I might add it real quick

• full being equivalent to verbosity=True
• minimal being equivalent to verbosity=False
• silent being equivalent to verbosity=False from current PR

I'd like to avoid magic string if possible. It's a bit unorthodox but I'm inclined to making it an Optional[bool] and use None/False/True so verbose=None (i.e. silent) is the default in transcribe() but from the command line it selects either True or False.

I can make these edits on this PR if you'd like/don't mind.

Don't bother yourself, I have nothing to do at the moment